CIGRE SCIENCE

& ENGINEERING
Volume No.13, February 2019

Innovation in the
Power Systems
industry
Enginers and specialists worldwide exchange
information and state-of-the-art world practices
to enhance knowledge related to power systems in
CIGRE’s latest publication.

"Best of" papers, Paris Session 2018

CIGRE 21, rue d’Artois, 75008 Paris – ISSN: 1286-1146

 Volume N°13, February 2019

Journal edited by CIGRE

President Rob STEPHEN
Vice President Technical Marcio SZECHTMAN
Secretary General Philippe ADAM / philippe.adam@cigre.org

Editorial Committee
Chief Editor Konstantin O. Papailiou
Editors Pierre Argaut Erli Figueiredo
Christian Franck Nikos Hatziargyriou
Xidong LIANG Pouyan Pourbeik

Rob Stephen
Copyright registration N° 013, February 2019
1286-1146 CIGRÉ
© Copyright CIGRE
Conseil International des Grands Réseaux Electriques
International Council on Large Electric Systems
21 rue d’Artois - 75008 Paris - France
Tel : 33 (0)1 53 89 12 90
http://www.cigre.org
"The author(s) of papers published in the CIGRE Science and Engineering Journal, are
granted permission to post a PDF copy of their paper, as published in the journal, for self-
archival purposes on their own personal websites, or similar outlets (e.g. researchgate.com).
This permission is granted with the following restrictions, requirements, and understanding:
1. CIGRE is the copyright owner of all articles. Please see the CIGRE copyright statement for
details. The articles cannot be disseminate for commercial use (i.e. for payment, etc.).
2. A clear reference must appear where it is posted stating that "The paper may also be
downloaded for free from the source CIGRE. Goto: http://www.cigre.org/Menu-links/
Publications/CIGRE-Science-Engineering. Then click to download the <Month/Year> Issue
of the CIGRE Science and Engineering Journal. The paper appears on pages <page numbers>.
3. The paper can only be posted on the author(s) website after it has been published and
released through the CIGRE website."

Submission Procedure https://www.cigre.org/GB/publications/cigre-cse

Dear readers,

Hope you have catched a good start in 2019 and are healthy and happy as ever. This is definitely
for us here at CIGRE Science & Engineering the case, as with this issue we are able to offer you
the “best of ” papers, one from each Study Committee, from the Paris Session 2018. This selection
is in my view a great distinction and it is probably one of the toughest selection procedures in the
industry.

Just for the record, out of over 1000 abstracts submitted to CIGRE Central office – please note that
most of them had to pass before at least one stringent selection within their National Committee –
and some 500 finally selected for the Session, the best three per Study Committee are proposed by
the SC Chair the Editorial Board of CSE, which then decides which one will finally be published.
This is about 2% hit rate and I would be eager to hear if any other scientific Journal can beat this
figure.

It also makes my life easier, as it remains for me only to congratulate the authors for this important
achievement and thank my editor colleagues and all persons involved in the production process of
our Journal for their hard work in preparing this issue.

Needless to say, I am convinced that you will enjoy reading it as much as I did!

Prof. Dr. Konstantin O. Papailiou

Chief Editor
konstantin@papailiou.ch

Cigre Science & Engineering • N°13 February 2019

3
Contents of this issue

page

Editorial 3
A1 - Investigations on ROCOF withstand capability on large synchronous generators 5
A2 - Uneven liquid flow distribution in radial ducts in transformer winding cooling systems
shown by CFD and experimental measurements 17
A3 - Optimizing the energizing scheme of unearthed shunt capacitor banks 26
B1 - Current and future applications of HPTE insulated cables systems 34
B2 - Comparative Study of the Long-term Reliability of HTLS Conductor Systems 45
B3 - Voltage uprating of existing Eskom high voltage substations when transient
voltage stress and available withstand strength are coordinated 53
B4 - Research and development of Ultra-High-Voltage VSC for the multi-terminal
hybrid ±800kV HVDC project in China Southern Power Grid 61
B5 - Power swing blocking and tripping - brazilian study case 66
C1 - Planning a meshed HVDC offshore grid in the North Seas 79
C2 - Icelandic operational experience of synchrophasor-based fast frequency
response and islanding defence 89
C3 - Use of avian radar as a method to study the effect of bird flight diverters 100
C4 - Measurement and analysis of harmonic data to assess the impact of installations
connected to high voltage systems 107
C5 - Smart TSO-DSO interaction schemes and ICT solutions for the integration
of ancillary services from distributed generation 115
C6 - Effectiveness of different flexibility options and innovative network technologies
for the use in the BDEW traffic light concept, on the basis of a German
distribution grid 124
D1 - Insulation characteristics in DC-GIS: Surface charge phenomena on epoxy spacer
and metallic particle motions 133
D2 - Benefit and resolution of operational issues for information and communication
systems using virtualization techniques in the electric power industry 143

Cigre Science & Engineering • N°13 February 2019

4
Investigations on ROCOF withstand
capability on large synchronous
generators
K. CHAN*, J. OESTERHELD, S. TEMTEM GE, Switzerland
J. HALDEMANN, Former GE, Switzerland
Summary
With the large integration of non-synchronous generation
into the electrical supply system, significant reduction
in system inertia is now a challenging issue for system
operators. One impact of system inertia reduction is
on the Rate of Change of Frequency (ROCOF), which
can reach a high value (> 6 Hz/s [1] depending on the
measurement duration [2] and severity of disturbance)
when a system disturbance occurs that involves either a
loss of generation or loss of power export such as a large
interconnector. ROCOF withstand capability requirements
are now appearing in many grid codes, typically requiring
generators to stay connected during high gradients of
grid frequency. However, such requirements are not
harmonized amongst the various national grid codes,
although frequency is shared within a synchronous area.
Usually only the maximum gradients (e.g. 1 Hz/s) are
defined, but the boundary conditions (such as duration
and frequency profile), methodology and principle in most
cases stay unclear and do not allow a detailed evaluation
from a generator capability point of view. In some extreme
case, such as in the Australian grid code, the generator is
required to withstand a ROCOF of up to 4 Hz/s for 0.25s.
Since international equipment standards do not currently
provide any guidance on ROCOF withstand capability, the
only way to determine such capabilities would be for the
generator original equipment manufacturer (OEM) to carry
out a detailed technical assessment for the unit for a specific
site. This requires detailed information to be available,
which can be a challenge particularly for older power
plants. Such assessments are very time intensive covering
electrical, mechanical, protection as well as control aspects
of the generating unit and other affected equipment in the
power plant. This paper describes the authors’ experiences
* kevin.chan@ge.com
KEYWORDS
Rate of Change of Frequency, system inertia, stability, withstand capability, grid frequency gradient, grid code.
Cigre Science & Engineering • N°13 February 2019
5
on performing ROCOF withstand capability investigations
for existing power plants in Europe as mandated by one
transmission system operator. Section 1 of this paper
provides an overview of the grid code requirements on
ROCOF and proposes a study methodology to ascertain
compliance to the ROCOF withstand capability. In Section
2, a grid model suitable for simulation of ROCOF events,
validation of the components of the power plant models
as well as simulation scenarios are presented. Section 3
provides the impact assessment on the major equipment
affected by ROCOF withstand capability. Finally, in Section
4, recommendations, lessons learnt and conclusions are
discussed.
1. Introduction
The rapid decline in recent years of system inertia in
the power system due to large scale integration of non-
synchronous renewable energy sources has resulted
in concerns by transmission system operators on high
ROCOF and the impact on conventional power generation
equipment [3,4]. Conventional power plant operators
have two main concerns aside from the ability to remain
synchronized. These are that a high ROCOF event will
cause a catastrophic failure of a unit – which is primarily
a safety concern for station staff – and that repeated high
ROCOF events will negatively impact the commercial life
of the plant. To ascertain and assess these risks caused by
high ROCOF events, currently the only option is to carry
out investigations on ROCOF withstand capability both for
existing and new build of conventional generation plants.
This paper describes the authors’ experiences on performing
ROCOF withstand capability investigations for existing
power plants in Ireland as required by EIRGRID and SONI
[3]. ROCOF withstand capability is not harmonized and
requirements differ significantly world-wide. However,
the methodology as described for determining the ROCOF
withstand capability can be applied regardless of the
requirements.
1.1 ROCOF withstand capability requirements
Current international electrical machine design standards do
not provide guidance on how to design large synchronous
generators to withstand high ROCOF. Furthermore, grid
code requirements on ROCOF withstand capability for
generating units vary significantly as these depends on the
characteristics of an entire synchronous area [5]. Some
examples of these ROCOF requirements stipulated in grid
codes are listed in Table I below.
grid codes, the need is arising to ensure that a systematic
approach is undertaken to ensure that a power plant is
compliant with the requirement. To-date only EIRGRID
and SONI, system operators of Ireland, have mandated
existing thermal power plants in their systems to carry out
investigations to determine ROCOF withstand capability,
under their Generator Studies Project [3]. The scope is
divided into two sets of studies:
I. Electrical dynamic simulations: The purpose of these
studies is to assist in assessing the oscillatory dynamic
interaction between the transmission system and generation
unit’s response to ROCOF values.

In the power industry, there is currently no common
consensus on how ROCOF should be measured, how
ROCOF capability should be defined or how the withstand
capability for power plant equipment can be validated. This
situation needs to be urgently addressed not only by the
system operators but also in conjunction with the power
industry’s professional institutions. Note that pre-normative
research has been established to develop new IEEE/IEC/
CENELEC standards for rate-of-change-of- frequency
(ROCOF) measurements in electricity networks.
1.2 ROCOF withstand capability assessment
methodology
With increasing number of system operators stipulating
ROCOF withstand capability in their respective country
II. Mechanical / plant integrity studies: This set of studies
aims to examine whether the safety or plant integrity issues
previously raised by some generators are material.
Based on the authors’ experiences in carrying out ROCOF
withstand capability investigations for existing power
plants in Ireland and Northern Ireland, an assessment
methodology is proposed as shown in Figure 1.
The two-phase approach has the advantage that if the
PHASE I investigations show minimum impact of ROCOF
events, PHASE II investigations could be avoided in part or
all together, thus reducing the cost of the investigations. If
the investigations are related to existing power plants, data
gathering can be very challenging as old power plants may
not have all the required technical information available

Table I: Examples of ROCOF withstand capability requirements in grid codes.

Country Requirements
Australia [7] Must be capable of continuous uninterrupted operation for frequencies defined unless the rate of change
of frequency is outside the range of –4 Hz to 4 Hz per second for more than 0.25 seconds.

Denmark [8] The plant must be able to withstand transient frequency gradients (df/dt) of up to
±2.5 Hz/s in the connecting point without disconnecting.
Europe [6] Requirement for Generator RFG, Article 13.1(b), non-exhaustive
Israel [9] >1.0Hz/s for 100ms; >0.5Hz/s for 300ms. Maximum rate of frequency changes shall be 3.0Hz/s.

Ireland [10] Remain synchronized to the Transmission System for a Rate of Change of Frequency up to and
including 1 Hz per second as measured over a rolling 500 milliseconds period.

Mexico [11] ROCOF limits: 2.5 Hz/s

Spain [12] ROCOF of 2 Hz/s from 750 ms to 300 ms.

Cigre Science & Engineering • N°13 February 2019

6
 Figure 1: ROCOF Withstand Capability – Assessment Methodology.
because an OEM may not exist anymore, upgrades have
been made using a different manufacturer’s equipment
or essential technical information for the ROCOF
investigation was not available or provided when the plant
was built. In such instances, best engineering assumptions
and/or additional testing on-site may have to be made. The
system operator must provide guidance as well as data
related to the ROCOF events to be investigated in their
system as this will determine how credible and reliable the
outcome of such investigations will be. The dynamic power
system study is the corner stone for initiating the ROCOF
withstand capability investigation as it will provide the
information for assessing the generation unit’s response as
well as providing the necessary electrical inputs (currents,
voltages, power, rotor angles, electromagnetic torques etc.)
for the assessment process. PHASE II investigations are
directed mainly to determine the impact of ROCOF events
on the mechanical design and construction of the rotating
equipment. The objectives of the tasks within PHASE I and
PHASE II investigations are given in detail in Sections 2
and 3 below. The deliverables of these investigations are
to provide:
1. A clear statement for the withstand capability of the
power plant according to the Grid Code requirements;
2. Recommendations of remedial upgrades, if any, are
necessary and possible.
Cigre Science & Engineering • N°13 February 2019
7
2. ROCOF simulation study
2.1 Power system simulation model
To ascertain the dynamic behavior of a power plant during
some ROCOF events, power system simulations must be
carried out. To assess the performance of a single power
plant within the power system, it is common to consider
the external power system as a single component whose
frequency can be varied with time to represent the frequency
during a ROCOF event. This grid representation approach
was used in the investigations carried out by the authors of
this paper. Additionally, the modeled system consists of the
following electrical components:
• Dynamic model of the generator with Automatic
Voltage Regulator (AVR) and Power System Stabilizer
(PSS);
• Full and equivalent lumped mechanical multi-mass
models;
• Turbine governor control systems;
• Transformers (Generator step-up and auxiliary);
• Auxiliary induction motors and composite static load;
• Medium Voltage Cables.
A typical system simulation model for ROCOF study is
shown in Figure 2.
The investigations consider various ROCOF events for
different power plant operating points. An example of this
 Figure 2: Typical system simulation model for ROCOF investigations.

is depicted in a workflow as proposed by EIRGRID [3] in
Figure 3. Each ROCOF event is represented by a unique
frequency trace shown as a system frequency profile over
a time interval.
2.2 Control system model validation
How the control systems of the generator respond to a
ROCOF event is critical to understanding the withstand
capability of the unit. To ensure the control behavior of
the simulation model is a true reflection of the physical
equipment, it is necessary to validate the performance and
behavior of the control system models. Validation in this
context entails the comparison of actual site tests carried
out on the unit with simulations using the model under
the same set of operating conditions. Both excitation and
turbine-governor control systems must be considered during
validation. The simulation models of excitation and turbine-
governor systems can be derived from those established in
several publications from IEEE and CIGRE [14,15,16].
In many instances, these models are available directly as
standard or user-defined models in the simulation software
being used. Figure 4 and 5 shows typical validation traces.
For the impact assessment of the generator and turbine
components as detailed in Section 3, relevant generator
state variables are required and these can be obtained
from the simulation results. Typical variables include the
following time series plots:
• Electromagnetic (air-gap) torque;
• Rotor electrical angle and speed;
• Active and reactive power;
• 3 phase stator voltages and currents;
• Excitation voltage and current.

Figure 3: ROCOF scenarios and frequency traces.

Cigre Science & Engineering • N°13 February 2019

8
 Figure 4: Validation of excitation model. Figure 5: Validation of turbine-governor model.

3. ROCOF impact assessment
3.1 Generator electrical design assessment
For the impact assessments on generator electrical design,
both the stator and rotor are considered. In general, for
the stator and rotor components, the impact of additional
thermal stresses due to increased currents, additional
electrical stresses due to increased voltages, as well as
increased electromagnetic forces that can occur during
a ROCOF event must be assessed. By careful study and
comparison of the results from all simulation cases, it is
possible to identify one or two worst-case scenarios. Key
indicators are operating points of the unit near full load
and close to the steady-state stability limits (i.e. extreme
leading power factor) as well as the unit experiencing
first rotor angle swings towards stability limit. The
following assessment are typical examples for a ROCOF
investigation. It starts with assessing the risk of transient
or rotor angle instability. This can be identified from a plot
of the generator rotor angle profile as shown in Figure 6.
For this ROCOF event (1 Hz/s) the rotor angle is less than
90° and has no over-shoot exceeding the angle stability
limit. However, for other more severe ROCOF events,
rotor angle instability may occur [17].

Figure 6: Generator rotor angle profile due to a 1 Hz/s drop ROCOF event.

At full active
power:
peak current
36kÂ
28kA
fluctuation

Figure 7: Stator current amplitude [A] vs. time [s] during the ROCOF event (blue) in comparison to the over-
current requirement of 1.5p.u. for 30s according IEC60034-3 (red).

Cigre Science & Engineering • N°13 February 2019

9
 Figure 8: Stator voltage [p.u.] vs. time [s] during ROCOF event.
Figure 9: Generator V/Hz [p.u.] vs. time [s] during ROCOF event.
*) inside limits of given V/Hz vs (t) diagram
The additional thermal stress to the stator windings due
to increased stator currents shall be checked against the
requirements of IEC60034-3 [18] with regards to short-
time over-current. IEC requires, that generators with rated
outputs up to 1200MVA shall be able to carry, without
damage, stator currents of 1.5p.u. for 30s and the thermal
equivalent loading duration as defined by the equation:
(I2-1)*t < (1.52-1)*30s =37.5s.
If the stator current is higher than 1.5p.u., the thermal
equivalent loading duration would correspondingly
be reduced. Figure 7, shows the current amplitude
corresponding to 1.9p.u. of rated stator current (In),
superimposed on the stator current amplitude versus time
plot during one ROCOF event. The comparison shows,
that the integral of 37.5s (content of red borders) during
the ROCOF event will never be reached and are within
the IEC design limits. The additional thermal stress to
the stator winding due to increased currents during this
ROCOF event is therefore considered as not critical. No
considerable impact on component lifetime is expected.
The additional electrical stress to the stator winding’s
insulation is due to increased voltages during a ROCOF
Cigre Science & Engineering • N°13 February 2019
10
event. IEC60034-3 stipulates a generator continuous
operating voltage range of ±5 % rated value. Figure 8
shows the simulated stator voltage amplitude profile for a
ROCOF event. Results show that the voltage fluctuations
do not exceed the voltage variation limit of 5%. Voltage
- frequency limits (over fluxing V/Hz) are as well not
exceeded as shown in Figure 9. The V/Hz permissible for
this unit is 1.15p.u. Hence, the electrical stresses experienced
under these conditions are evaluated as not being critical
for the insulation systems, due to the short duration of the
overvoltage. No substantial impact on component lifetime
is expected.
The impact of increased electromagnetic forces to the end
windings of the generator stator can be assessed using a
3D end winding model (Figure 10) to calculate the end
winding forces considering the stator winding, the rotor
winding, the press plate and housing boundary. The model
is derived from mechanical design and the electromagnetic
calculation was carried by the Biot-Savart method. The
phase ring arrangement is based on a simplified geometry
fitting to the electromagnetic tool used. The 3D results
condensed into a 2D view regarding 3-axes for rated current
is shown in Figure 11.
 Figure 10: 3D end winding model for end region field and force calculation. Figure 11: Results of the 3D end winding force calculation in 10° sections
in a 2D view for rated current. Radial forces (shown in BLUE), Tangential
forces (in MAGENTA), Axial forces (in GREEN).

From static forces point of view the expectable risk can be
estimated by comparing the currents and corresponding
forces during a ROCOF event with the forces expected
in a 3-phase short circuit the generator is designed for. As
an example, the maximum 3-phase short circuit current
(peak value) is calculated with a simplified equation using
saturated sub-transient reactance x”dv from the generator
data:
For comparison, in the analyzed ROCOF events a
maximum current of 1.85p.u. is reached. As forces change
proportional to the square of the currents, this would give a
fault force relation of
Thus, from static forces point of view the electromagnetic
forces to the end winding basket of the stator during a
ROCOF event appear not critical. They are sufficiently
below the values considered for machine design
during faults like 3-phase sudden short-circuits, mal
synchronizations, short-circuits in grids. The impact of
additional thermal stress to the rotor windings during
an ROCOF events is evaluated for the condition of field
current (If) reaching ceiling values. It must be ensured
that this value does not reach the maximum duration
allowable for ceiling operation as designed, typically 10s
[19]. As an example, the rated field current (Ifn) is 4347A
with a designed ceiling factor of Ifmax/Ifn=1.6. Figure
12 shows the field current profile. As the field current
neither exceeds the ceiling value nor the allowable ceiling
duration, the thermal stress to the rotor winding is not
critical and no impact on component lifetime is expected.
The impact of additional electrical stress to the rotor
winding’s insulation during an ROCOF events is evaluated
for the condition of field voltage (Uf) reaching ceiling
values for the allowable duration. For this example, the
rated field voltage (Ufn) is 374V with a designed ceiling
factor of Ufmax/Ufn=2.0. Figure 13 shows the field
voltage profile. As the field voltage neither exceeds

Figure 12: Field current profile. Figure 13: Field voltage profile.

Cigre Science & Engineering • N°13 February 2019

11
 Figure 14: Generator rotor winding showing magnetic field and current directions.

the ceiling value nor the allowable ceiling duration, the
electrical stress to the rotor winding insulation is not
critical and no impact on component lifetime is expected.
Increased electromagnetic forces during a ROCOF
event have an impact on the stress to the end winding
of the generator rotor. To evaluate the electromagnetic
(radial) forces during a ROCOF event it was considered
practical to compare these forces with the centrifugal
forces on the rotor end winding. For the calculation of
the electromagnetic forces a simplified approach was
applied and the following was assumed:
• Rotor field orthogonal to the conductors and parallel
to retaining ring (retaining ring forms a screen for
the rotor field assuming all rotor current is a transient
current which is a pessimistic assumption);
• Consequently, this field distribution leads to radial
forces on the rotor winding (radially inwards);
• The level of these electromagnetic forces can be
calculated by the basic formula F = B x I, where B is
the magnetic field and I being the current flowing in
a conductor;
• For a worst-case scenario, the forces are calculated
as applicable in the rotor slot;
• The rotor current maximum for the case of full load,
over excited has been taken to calculate the electro-
magnetic forces.
Figure 14 shows the magnetic field and current directions
in the generator rotor winding. As generators are designed
to withstand forces during faults like 3-phase sudden
short-circuits, mal synchronizations, short-circuits in grids
etc., comparison of the calculated values obtained from
the proposed method above, can be used to assess whether
the electromagnetic forces experienced by the rotor end
windings during this ROCOF event is critical or not as
well as if there is any impact on component lifetime.
3.2 Generator electrical protection assessment
An important consideration for ROCOF withstand
capability of the generator is to determine if any
generator electrical protection function will trip the
unit. Hence a key objective is to analyze the protection
functions behavior (picking-up and generating alarm or
trip signals) in case of ROCOF events. Typical protection
functions to be assessed are:

- 21 Minimal impedance - 24G Generator Overexcitation

- 27 Undervoltage - 32 Reverse power
- 40 Loss of excitation - 46 Negative phase sequence
- 49G Stator winding temperature - 51V Voltage restrained overcurrent
- 59 Overvoltage protection - 78 Pole slip
- 81 Underfrequency - 81.4 Overfrequency

Cigre Science & Engineering • N°13 February 2019

12
 Stage 1 Stage 2 Stage 3
Case study Pick-up Alarm/Trip Pick-up Trip Pick-up Trip
21 Minimal impedance 0 0
24G Generator overexcitation P 0
27 Undervoltage P 0
32 Reverse power N/A N/A 0 0
40 Loss of excitation 0 0 0 0
46 Negative phase sequence 0 0 0 0
49G Stator winding temperature P 0 0 0
51V Voltage restrained overcurrent P 0
59 Overvoltage 0 0 0 0
78 Pole slip 0 0
81 Underfrequency P A 0 0 0 0
81.4 Overfrequency 0 0
Table II: Summary of protection function behavior assessment for Pick-up, Alarm or Trip.

These protection functions can be modeled in many
simulation software (e.g. MATLAB SIMULINK, EMTP-
RV). Table II provides an example of the assessment of the
generator protection functions.
The generator overexcitation protection function is picking-
up but the pick-up time is short enough that no alarm or trip
action is generated. The undervoltage protection function
is picking-up but the pick-up time is short enough that no
trip action is generated. The stator winding temperature
protection function is picking-up but the pick-up time is
short enough that no alarm or trip action is generated. The
voltage restrained overcurrent protection function is picking-
up but the pick-up time is short enough that no alarm or trip
action is generated. The underfrequency protection (Stage
1) is picking-up and an alarm is generated. The duration of
the alarm is 0.24 seconds. Figure 15 shows the behavior of
the underfrequency protection for this ROCOF event.
3.3 Shaft line lifetime assessment
ROCOF events could hypothetically, if sufficiently severe,
result in reduced component lifetime, decreasing overhaul
intervals, increased inspection requirements, or in the worst
case, catastrophic failure. This could be due to either a single
ROCOF event or cumulative damage caused by a series
of such events. Hence, a lifetime assessment of the shaft
line, due to high transient torque levels during a ROCOF
event, should be carried out, if deemed necessary. The
electromagnetic torques in the air gap of the generator are
those calculated in Section 2. These torques are imported
into the rotor dynamics calculation tools. Based on this

Figure 15: Behavior of the underfrequency protection showing Pick-up and Alarm Signals.

Cigre Science & Engineering • N°13 February 2019

13
 Figure 16: Calculation model used for assessments.
model, a torsional analysis will be performed identifying
the highest loaded part of the shaft line. Such torsional
analyses as described would consist of two parts:
a) Eigenfrequency calculation and b) torsional forced
vibration analysis. Due to the proprietary nature of shaft line
models it is not possible to provide any example figures or
analysis in this paper. Suffice to say, for the investigations
carried out, the stresses during the ROCOF events were
well below the design endurance limits of the shaft line and
no additional lifetime consumption is expected.
3.4 Generator stator core and casing foundation
lifetime assessment
For the mechanical assessments of the generator stator the
following calculations are performed:
1. Stresses at the core suspension;
2. Forces acting on the generator stator casing foundation
and to the foundation bolts.
The calculations of required core suspension stresses and
forces at generator foundation are performed for the worst
ROCOF event with regards to generator electromagnetic
(air-gap) torque. The weakest components of the stator core
suspension are the screws that are clamping the support
plates to the stator core and to the circumferential ribs of
the housing. By calculating the shear stresses on the screws,
the corresponding fatigue limit can be determined and the
impact on component lifetime can be assessed.
The vertical forces acting on the generator foundation due to
ROCOF must be superimposed with the gravity load acting
on the complete generator stator and rotor. This will yield
the maximum compressive load acting on the generation
foundation. In our experience, no tensile loads are acting on
the foundation and bolts due to ROCOF events.
The mechanical assessments in this example shows no
relevant impact on the lifetime of the stator core suspension
system or the foundation bolts.
Cigre Science & Engineering • N°13 February 2019
14
3.5 Generator rotor retaining ring lifetime assessment
Typically, 3D finite element (FE) calculation is performed
to determine the stresses in the retaining ring, retaining ring
cover and rotor tooth. The calculation model consists of
rotor, retaining ring and cover plate. Symmetry boundary
conditions are applied on symmetry planes. For nominal
operation, rotational speed (rated speed) and pressure due
to winding below retaining ring are applied to the model.
The nominal operation is first considered in the calculation
to obtain the stress level of retaining ring. In the second
step, only angular acceleration obtained from rotordynamic
calculation is applied to the model. By comparing the stress
levels for both cases, an impact of ROCOF events can be
assessed. Figure 16 shows the calculation model used for
the assessment. From the investigations in this example,
the angular acceleration has an insignificant impact on the
stress levels of retaining ring, hence, no impact on lifetime
of retaining ring is expected. For other rotor components
that are less loaded than the retaining ring, the same result
in regards of lifetime is expected.
3.6 Generator excitation system assessment
The PSS output signal tends to saturate at its upper limit
when the ROCOF is positive, and to its negative limit when
the ROCOF is negative. This PSS saturation modifies the
AVR reference in such a way that the generator terminal
voltage is increased above its initial value for positive
ROCOFs (positive PSS saturation), and reduced for
negative ROCOFs (negative PSS saturation). It should be
noted that PSS behavior can have a negative impact during
fast ROCOF events under certain loading conditions and
the PSS disconnection logic implementation. Hence it
is necessary to investigate this for all credible loading
conditions as well as the implemented PSS control logic.
The underexcitation limiter (UEXCL) is activated in all
leading power factor cases due to the operation of the PQ
limiter. The overexcitation (OEL) limiter is not activated.
The threshold of V/Hz limiter is not reached in the simulated
cases but higher ROCOF values may trigger this limiter.
3.7 Generator turbine-governor system assessment
The turbine-governor set responds to the fast frequency
change and should contribute correctly by increasing power
for frequency drop and decreasing power for frequency rise
profiles, the response is more significant after the first 2 s of
the frequency transients (change from nominal frequency
to approximately +51/-48.5 Hz, depending on the profile).
The overall response of the unit is mostly due to the action
of the speed droop control function; the frequency response
functions and turbine dynamics are slow acting compared
to the transient applied to the unit and associated electrical
grid interaction. Although the investigations carried out
by the authors did not provide any evidence for concerns,
some areas with regards to gas turbine combustion
and operability may require further assessment. These
assessments could be in the areas of gas combustion lean
blow out (LBO) risk and compressor operating limit line
(OLL) infringement. These areas could be potential issues
for ROCOF withstand capability for gas and combined
cycle plants. In our experience, governing and other critical
control simplifications could also result in a study missing
potential issues.
4. Conclusion
This paper describes the authors’ experiences on performing
ROCOF withstand capability investigations for existing
power plants in Ireland as required by EIRGRID and SONI.
An overview of the grid code requirements on ROCOF
withstand capability was provided. These requirements are
quite wide ranging and are not harmonized. Furthermore,
international standards do not provide guidance or
recommendations on how to design a synchronous machine
to ride through a ROCOF event. Currently the only way of
ensuring a power plant is ROCOF compliant is to conduct
a comprehensive engineering investigation of critical
components of the power plant. A study methodology has
been described in Section 2, proposing what to assess for
ROCOF withstand capability. Section 3 details the actual
process of impact assessment of the various components of
the generating unit focusing on assessment procedures. The
lessons learnt are:
• The investigation is first of its kind, so the scope
was difficult to define due to many unknowns. Data
Cigre Science & Engineering • N°13 February 2019
15
collection is time consuming and available technical
information on older machines may be an issue. Such
investigations involve interdisciplinary teams and
require substantial engineering resources and time for
completion;
• The study cases provided do not necessarily represent
all the situations the equipment will be exposed
to, considering the changing grid and generation
resources. Grid voltage at nominal values, as stipulated
by the grid operator for these investigations, might not
result in the worst- case scenario;
• Simplified single bus grid model versus detailed full
grid model for ROCOF studies is to be understood
further;
• PSS behavior is to be investigated further for such
events;
• Instability of the unit depends on the magnitude of
ROCOF values as well as grid parameters such as
short-circuit level and the unit’s operating conditions;
When pole-slipping occurs, safety and equipment
damage should take precedent over the need to keep
unit connected to the grid;
• The mechanical stress experienced by a power plant
as per 3-phase fault or mal- synchronization events are
several times higher than that observed in a ROCOF
study. These events are rare and considered in the
power train design standards. However repetitive
ROCOF events as well as associated stress are not
considered in initial design or lifetime assessment and
may need to be monitored;
• Though not reported in this paper, investigations were
also carried out on the impact on the balance of plants
electrical equipment. No critical issues were identified.
Regarding the combined variation of voltage and
frequency, the performance of motors connected to
the auxiliary board is within zone B of IEC 60034-1
which is acceptable for continuous operation;
• Compliance will be demonstrated by injection tests on
site. Additional data to be collected to assess ROCOF
withstand capability;
• The authors’ approach validated the need for a detailed
gas turbine and parts of a plant simulation. Multiple
items that could impact connection status were
identified in the first phase of work;
• The investigations want to clearly define issues
found in two categories. One where there may be a
 disconnection, or non-compliance and the other where
there may be an equipment impact.
In case of non-compliance, the Generator owner is to apply
for derogation and provide remedy for non-compliance.
In addition, the generating unit will not be rewarded for
ROCOF compliance but instead be penalized and would
not be on the priority list for unit dispatch.
In the authors’ opinion, international standardization
organizations and transmission operators should pool
their resources to harmonize ROCOF withstand capability
definition as well as to provide guidance on how to
show proof for ROCOF compliance in an efficient and
cost-effective manner for both existing and new build
conventional power plants.
The authors would like to thank the GE team for their
invaluable contributions to these ROCOF investigations.
5. Bibliography
[1] AEMO, “BLACK SYSTEM SOUTH AUSTRALIA 28
SEPTEMBER 2016 – FINAL REPORT”, 23 March 2017.
[2] EIRGRID, SONI, “Summary of Studies on Rate of Change of
Frequency events on the All-Island System”, August 2012.
[3] EIRGRID, SONI, “Ireland ROCOF Generator Project - Study
Cases for Electrical Dynamic Simulations”, 23 December 2014.
[4] National Grid, UK, “System Operability Framework”, November
2015
[5] ENTSO-E, “Rate of Change of Frequency (ROCOF) withstand
capability - Implementation Guidelines”, 16 November 2016.
Cigre Science & Engineering • N°13 February 2019
16
[6] EC, “Establishing a network code on requirements for grid
connection of generators”, 14 April 2016.
[7] AEMC, “Australia National Electricity Rules Version 95”, 15
August 2017.
[8] ENERGINET, Denmark, “Technical regulation 3.2.3 for thermal
plants above 11 kW”, 10 January 2017.
[9] Israel Electrical Company, “Super-voltage Connection of
Customers”, 2 March 2009
[10] CER, Ireland, “Rate of Change of Frequency (ROCOF)
Modification to the grid code”, 4 April 2014.
[11] CRE, Mexico “Technical Requirements Manual for the
Interconnection of Electric Plants to the National Electric System”,
8 April 2016.
[12] RED, Spain, “Desarrollo de los requisitos técnicos del Reglamento
(UE) 2016/631”, June 2017 [13] CIGRE Working Group A1.29,
“Guide on New Generator-Grid Interaction Requirements”, 2018.
[13] CIGRE Working Group A1.29, “Guide on New Generator-Grid
Interaction Requirements”, 2018.
[14] IEEE Std. 421.5-2016, “IEEE Recommended Practice for
Excitation System Models for Power System Stability Studies”, 15
May 2016.
[15] CIGRE Technical Brochure 238, “Modeling of Gas Turbines and
Steam Turbines in Combined Cycle Power Plants”, April 2003.
[16] IEEE Task Force Technical Report PES-TR1, “Dynamic Models
for Turbine-Governors in Power System Studies”, January 2013.
[17] DNV KEMA, “ROCOF, An independent analysis of the ability of
generators to ride through rate of change of frequency values up to
2Hz/s”, 8 February 2013.
[18] IEC 60034-3, Ed. 6, “Rotating electrical machines – Part 3:
Specific requirements for synchronous generators driven by steam
turbines or combustion gas turbines”, November 2007.
[19] IEEE Std. C50.13-2014, “IEEE Standard for Cylindrical-Rotor
50 Hz and 60 Hz Synchronous Generators rated 10 MVA and
above”, 27 March 2014.
Uneven liquid flow distribution in radial
ducts in transformer winding cooling
systems shown by CFD and experimental
measurements
P. JARMAN*, National Grid, UK
X. ZHANG, M. DAGHRAH, Q. LIU, Z.D. WANG, The University of Manchester, UK
P. DYER UK, Power Networks, UK
A. GYORE, M&I Materials, UK
P. SMITH, Shell Global Solutions, UK
P. MAVROMMATIS, TJH2B, UK
M. NEGRO, Weidmann, Switzerland
D. WALKER, Scottish Power, UK

Summary PIV measurements that liquid flow distribution in the

Determination and optimisation of the winding hot-
spot temperature is essential for managing the loading
capability and thermal ageing of transformers. The
liquid flow in the various winding cooling ducts directly
impacts the position and temperature of the hot-spot. This
paper uses experimental measurements using particle
image velocimetry (PIV) on a test rig and computational
fluid dynamics (CFD) to model the flows in a zig-zag
type directed flow disk type winding cooling system to
show how the flow distribution, hot-spot position and
temperature varies with flow rate. Traditional thermal-
hydraulic network models which are commonly used to
predict flows and hot spot temperatures do not predict
some of the phenomena observed, such as reverse flow at
high and low flow rates and consequently have a restricted
range of flow rates where they produce accurate results.
These results are very important for good thermal design
and this research shows for example, that increasing
already high liquid flow rates does not necessarily lead to
reduced temperatures.
The research was conducted at Manchester University,
funded and guided by a transformer research consortium
comprised of the authors’ companies.
For higher flow rates typical of pumped oil directed (OD)
cooling modes, it is found by both CFD simulations and
winding is governed by the winding cooling duct geometry
and the winding inlet flow rate (determined by the pump),
while heat transfer to the liquid from the winding has a
negligible influence on the flow distribution. Therefore,
the hydraulic system and the thermal system can be
decoupled. Dimensional analysis is used to investigate
the effects of winding geometric dimensions, liquid types
and total liquid flow rates on the liquid flow distributions.
It is found that the dimensionless flow distribution, i.e.
the volumetric flow fraction in each radial (horizontal)
cooling duct is mainly controlled by the Reynolds number
(Re) at the winding inlet and the ratio of radial duct height
to axial duct width (α). The higher Re (i.e. higher liquid
inlet velocity or higher liquid temperature) or higher α, the
more uneven the flow distribution will be. When Re or α
is high enough, stagnation and reverse flows at the bottom
part of the winding pass can occur. PIV measurements of
the flow distribution in a physical disc-type winding model
with a range of geometries verify quantitatively the CFD
results, the flow distribution trend and the occurrence of
reverse flows.
For low flow rates typical of oil natural (ON) cooling
modes, buoyancy forces are responsible for the liquid
circulation between the windings and the radiators. The
hydraulic system and the thermal system are fully coupled
and therefore cannot be considered separately. The

*paul.jarman@nationalgrid.com

KEYWORDS
Transformer, thermal ageing, hot-spot temperature, OD, ON, reverse flow, CFD, PIV

Cigre Science & Engineering • N°13 February 2019

17
most uniform flow distribution is found to be under the
condition of balanced buoyancy forces and inertial forces.
Both CFD simulations and PIV measurements show that
flow distributions feature a pattern of higher flow rates
in the bottom part of the passes and lower flow rates in
the top part of the passes, opposite to that seen at higher
flow rates. The occurrence of stagnant or reverse flows
in the top winding passes is found under the condition of
low total liquid flow rate or high total power loss in the
winding. With a further decrease of the total liquid flow
rate or increase of the total power loss in the winding, the
region suffering from reverse flows will shift down to the
middle ducts of the passes. The occurrence of stagnant or
reverse flow is found to result in a significant increase of
the hot-spot temperature.
The results presented in this paper explain certain
overheating phenomena observed in real transformer
windings following failure and provide a useful
background to the review of transformer winding cooling
flow designs.
1. Introduction
This research was initiated and funded by the consortium
of companies indicated in the author list as a direct
result of observations made on scrapped transformers
indicating that overheating had occurred in the windings
to an extent and in positions that would not have been
predicted by traditional estimation methods for example
those assuming even flows and a calculated winding loss
distribution. The authors desired to have an understanding
of the oil flow distribution in typical transformer winding
geometries for use as a design review tool as well as for
failure analysis and transformer rating calculations. To
do this it was considered necessary to validate computer
modelling techniques against actual flow and temperature
measurements made in a representative test rig.
To cover one of the most common winding cooling
geometries, the investigation examined the liquid flow
distribution in the radial ducts of a zig-zag type cooling
system of a disk type transformer winding. To cover the
range of designs and operational regimes, the liquid flow
distribution and its impact on temperature distribution
was investigated for both high flow (typical of a pumped
OD system) and low flow (typical of a thermosiphon ON
cooling mode) conditions.
Cigre Science & Engineering • N°13 February 2019
18
After consideration of several possible techniques, flow
measurements on a test rig were made using particle image
velocimetry (PIV) and the computer model employed
computational fluid dynamics (CFD) simulations.
Dimensional analysis has been adopted to guide the range
and reduce the number of experimental measurements
and CFD simulations required. The significance of
some dimensionless numbers that emerged from the
dimensional analysis is demonstrated and emphasised.
Liquid flow distribution in the radial cooling ducts has a
direct impact on the cooling performance, and a relatively
uniform flow distribution is desired to avoid localized
overheating. The uniformity of liquid flow distribution is
controlled by the winding geometry, liquid properties and
the liquid flow rate in the winding.
2. Methodology
To simplify the relationship between flow distribution
and its controlling parameters without losing accuracy,
dimensional analysis has been adopted. Dimensional
analysis is performed prior to obtaining quantitative
answers to the problem and it provides simplicity, insight
and universally applicable solutions [1].
When the flow is driven by the pump, dimensionless
flow distribution, indicated by volumetric flow fractions
in radial ducts, is found to be controlled by the Reynolds
number (Re) at the winding pass inlet, which provides a
measure of the ratio of inertial forces to viscous forces
[1]. When the flow is mainly driven by buoyancy forces,
dimensionless flow distribution is found to be mainly
controlled by the Richardson number (Ri), which provides
a measure of the ratio of buoyancy forces to inertial forces
[2].
Experimental test rigs of disc-type windings were also
built to measure flow and temperature distributions
under various operational conditions. A particle image
velocimetry (PIV) system was used to measure liquid
flow rates in radial cooling ducts, and thermocouples
were used to measure temperature distribution in the
winding models. 2D CFD models of disc-type transformer
windings were established using commercial software,
COMSOL Multiphysics. From the CFD models, detailed
flow and temperature distribution information can be
obtained.
 Fig. 1. Schematic diagram of the experimental setup. (a) Circulation loop. (b) Aluminium plate with cartridge heaters and thermocouples.[3]

3. Experimental set-up captured. Therefore, based on the theoretical parabolic

velocity profile, the average velocity in the duct can be
The experimental set-up is shown in Fig. 1 (a) [3] calculated by dividing the maximum velocity by 1.5. The
and consists of a series of heated plates arranged in a flow distributions reported were measured in pass 3 and
rectangular section clear plastic tube construction to the radial ducts are numbered from 1 at the bottom to 11
represent a portion of a disc-type winding, a radiator, a at the top, as shown in Fig. 1 (a).
main tank, an expansion vessel, a flow meter, a heating
unit to control winding inlet oil temperature, fittings
and pipes. The winding model comprises 3 passes with
10 aluminium plates and 11 radial ducts per pass. Each
aluminium plate has a dimension of 100×104×10 (mm)
and is fitted into 3 mm grooves on the two side walls to
make the radial duct depth 94 mm and width 104 mm
[3]. Two cartridge heaters are embedded in each plate, 30
mm from the two edges of the plate, making the distance
between the two heaters 44 mm [4]. Two thermocouples
are fitted in the middle of each plate 10 mm from the two
edges, as shown in Fig. 1 (b).
Particle image velocimetry was used to measure the liquid
flow rates in the radial (horizontal) cooling ducts. The
mechanism of the PIV system is shown in Fig. 2. Seeding
particles were added to the liquid; these particles disperse
evenly and follow the flow faithfully. Two laser pulses
with a pre-set time interval between them were used to
illuminate the seeding particles in the radial cooling Fig 2. PIV system with an acquired experimental example of the parabolic
ducts. Each time the laser fires, the synchroniser triggers velocity profile in a radial cooling duct.[4]
the camera to take an image so a pair of images (frame
A and frame B, as shown in Fig. 2 bottom left) is taken 4. Investigation of high flow rates
for the two consecutive laser pulses. Multiple image pairs High flow rates driven by the pump simulate an oil
are generated for each PIV measurement, and then PIV directed (OD) cooling mode. Power loss was set to 50
software analyses the image pairs statistically to infer W/plate. It has been found that the power losses in the
the velocity profile in the radial cooling duct. A typical winding do not alter liquid flow distribution [1]. Therefore,
measured velocity profile is shown in Fig. 2 bottom right. isothermal flows are representative at these flow rates.
Since flows in the radial cooling ducts are laminar, a From dimensional analysis, it is found that for a fixed
quasi-parabolic velocity profile is obtained. The velocities winding geometry, the flow distribution in the winding
very close to the duct wall are not properly captured due pass is mainly controlled by Re at the pass inlet and the
to strong light reflections, whereas the velocities in the ratio of radial duct height to axial duct width (Hduct/Wduct as
middle part of the radial cooling duct are accurately shown in Fig. 1 (a)), referred to as α [1].

Cigre Science & Engineering • N°13 February 2019

19
4.1 Verification of Dimensional Analysis for Flow The measured flow distributions of the three cases are
Distribution shown in Fig. 3. The average velocity distributions are
different due to different total liquid flow rates, as shown
Experimental tests were conducted to verify the conclusion
in Fig. 3 (a), whereas the volumetric flow proportion
that by matching Re the dimensionless flow distribution
distributions are almost identical, proving the conclusion
will also be matched [1]. The Reynolds number is defined
from dimensional analysis.
as:
4.2 Flow Distribution for a Range of Re
Re =
A parametric sweep of Re is needed to identify the
quantitative relationship between the flow distribution and
where um for average pass inlet velocity (m/s), Wduct for Re. It would be difficult to perform a large enough number
axial duct width (m), ν for kinematic viscosity (m2/s). of experimental tests to establish this relationship, so 2D
Since the kinematic viscosity of oil decreases with CFD simulations were performed instead. Dimensionless
temperature, a lower oil flow rate can be compensated for by flow distribution results from CFD simulations have
a higher oil temperature to maintain the same Re. To test this, been extracted and correlated to Re to form a predictive
three cases shown in Table 1 were measured on a benchmark correlation equation.
winding model with radial duct height 4 mm and axial duct Experimental tests, shown in Table 2, were conducted
width 10 mm. The liquid used was mineral oil. to cover the practical range of Re for high flow (OD)
cooling modes and to verify the validity of the correlation
Table 1. Three cases with similar Re [3] (using mineral oil)
equation. Comparisons of flow distributions from PIV
Case 1 Case 2 Case 3
measurements and those from the correlation equation are
shown in Fig. 4. It can be seen that the calculated and
Pass inlet oil flow rate (litre per minute) 6 12 18 measured results match closely and that with an increase
Pass inlet velocity (m/s) 0.1 0.2 0.3
of Re, the flow distribution gets increasingly uneven.
What is more, for case 5 (highest flow Re = 1402), reverse
Oil temperature ( ° C ) 77 48 36 flow in duct 1 at the bottom of pass 3 was shown by both
Re (Wduct=10 mm) 541 536 547 the PIV measurement and the correlation equation derived
from CFD simulations.
Table 2. Three cases to cover the practical range of Re [3]

Case 4 Case 5 Case 6

Pass inlet oil flow rate (litre per minute) 6 12 18

Pass inlet velocity (m/s) 0.1 0.2 0.3

Oil temperature ( ° C ) 20 48 70
Re (Wduct=10 mm) 97 536 1402

Fig. 3. Comparison of flow distributions for three cases in Table 1 with similar
Re. (a) Average velocity. (b) Flow proportion where the total oil flow rate is
regarded as one unit. [3]

Cigre Science & Engineering • N°13 February 2019

20
 Fig. 4. Comparison of flow distribution for Re in a practical range for OD Fig. 5. Comparison of flow distribution for two cases of different horizontal duct
cooling modes with vertical duct width being 10 mm. (a) Average velocity in heights. (a) Average velocity in each duct of pass 3. (b) Flow proportion in each
each duct of pass 3. (b) Flow proportion in each duct of pass 3 where the total oil duct of pass 3 where the total oil flow rate is regarded as one unit. The legend
flow rate is regarded as one unit. The legend ‘equ’ refers to correlation equation ‘equ’ refers to the correlation equation derived from CFD parametric sweeps.[3]
derived from CFD simulations. [3]
4.4 The Effects of Re and α on Flow Distribution
4.3 The Effect of α (radial duct height divided by
Uniformity
axial duct width) on Flow Distribution
The uniformity of the flow distribution in the winding pass
The ratio of radial duct height to axial duct width, α, is
can be characterised by a parameter called maldistribution,
identified by dimensional analysis and CFD parametric
defined as the ratio of the maximum radial duct flow
sweeps to influence the flow distribution in the radial
velocity to the minimum radial duct flow velocity in
ducts. The comparison of flow distributions for two cases
the winding pass. The variation of the maldistribution
with different α but with the other parameters kept the
against Re and α obtained from the correlation equation
same (see Table 3) is presented in Fig. 5. It can be seen
is shown in Fig. 6. It can be seen that the maldistribution
that the calculated results from the correlation equation
increases with increasing Re and α monotonically. If
are close to those from PIV measurements and in the case
a maldistribution of 5 is set to be the upper limit, then
of high α (narrow axial ducts), the flow distribution is
region 1 shown in Fig. 6 with relatively low Re and low α
more uneven and reverse flow at the bottom of the pass is
is permissible, whereas region 2 with high Re and high α
both predicted and observed.
is not. It is worth noting that α is only shown from 0.25 to
Table 3. Two cases with different α’s and identical other parameters
0.33, because in the region of higher α and high Re (e.g.
1200) flow stagnation or reverse flow can occur making
the maldistribution infinite or negative, which makes the
Case 6 Case 7
trend of maldistribution against Re and α difficult to plot.
Pass inlet oil flow rate (litre per minute) 12 12
Pass inlet velocity (m/s) 0.2 0.2

Oil temperature ( ° C ) 70 70
Vertical duct width (mm) 10 10
Horizontal duct height (mm) 4 6
α 0.4 0.6

Re (Wduct=10 mm) 878 878

Cigre Science & Engineering • N°13 February 2019

21
 CFD simulation is 3 ºC. This could be because of strong
3D flow phenomena under reverse flow conditions. 3D
CFD simulations are being conducted to check the effects
of 3D flow phenomena on temperature distribution.

Fig. 6. Variation of maldistribution with Re and α. In region 1, maldistribution is

smaller than 5, whereas in region 2 maldistribution is larger than 5. [3]

Fig. 7. Comparisons of temperature and flow distributions between

4.5 Effect of Flow Distribution on Temperature measurements and CFD simulations with pass inlet oil velocity being 0.2 m/s.

Temperature and flow measurements were made on the

benchmark winding model with a uniform heat injection
of 50 W/plate and the inlet oil temperature controlled to be
70 ºC. The plate temperatures in pass three were obtained
by averaging the measurements of the two thermocouples
in each plate. 2D CFD simulations were made with
identical input conditions to the tested ones.
The comparisons of temperature and flow distributions
for a pass inlet velocity of 0.2 m/s are shown in Fig. 7.
For this flow rate, the simulated and measured flow and
temperature distributions are very close with a maximum
observed temperature difference of 1 ºC.
Fig. 8. Comparisons of temperature and flow distributions between
When the pass inlet velocity is increased to 0.27 m/s, the measurements and CFD simulations with a pass inlet oil velocity of 0.27 m/s.
flow distribution gets more uneven and the occurrence of
reverse flow at the bottom of the pass is shown by both 5. Investigations at low flow rates
PIV measurements and CFD simulations, as shown in Experimental tests of low total oil flow rates were
Fig. 8. Comparing the two cases it can be seen that as conducted for the benchmark winding model (radial
the inlet velocity increases from 0.2 m/s to 0.27 m/s duct height 4 mm, axial duct width 10 mm) to simulate
the measured highest temperature increases by 4 ºC. In oil natural (ON) cooling modes. For these low flow
this scenario at least, cooling performance is reduced tests, the heat input was reduced to 30 W/plate. The
rather than enhanced by a higher total liquid flow rate. winding pass inlet velocities, Reynolds (Re), Grashof
Comparing the flow distributions for the two cases, one (Gr) and Richardson (Ri) numbers are shown in Table
can notice that flow rate in ducts 9 and 10 from PIV 4. The Richardson numbers (Ri) are all larger than one,
measurements are consistently higher than the CFD indicating that buoyancy forces are significant in the tests.
simulation results. This is likely to be due to geometric Temperature distributions and flow distributions for these
deviations. For this higher flow rate case, the maximum ON tests are shown in Fig. 9. It can be seen that when Ri is
temperature difference between measurements and the 2, temperature and flow distributions are quasi-uniform,

Cigre Science & Engineering • N°13 February 2019

22
while with an increase of Ri the flow distribution gets winding. Different winding models can be built to better
increasingly uneven with higher oil flow fractions at the represent designs optimised for OD and ON cooling
bottom part of the winding pass. When Ri is high enough, modes. A large number of CFD simulations can be, and
reverse flows start to occur at the penultimate radial duct have been, made to cover a wide range of conditions.
at the top of the winding, as shown with an inlet velocity 6.1 OD Cooling Modes
of 0.017 m/s. CFD simulations show that with a further
increase of Ri the region suffering from reverse flow shifts The geometry of the OD winding model is shown in Fig.
downwards to the middle part of the pass and flows in the 10. The axisymmetric 2D CFD model has 3 passes with
top part of the pass recover to the normal flow direction. 6 discs per pass. Each disc comprises 30 paper wrapped
It is worth noting when the inlet velocity is increased conductors with a total width of 3 mm and a height of 12
from 0.017 m/s to 0.033 m/s, the highest temperature is mm. The width of the axial ducts is 8 mm, and the height
decreased from 93 ºC to 58 ºC. of the radial ducts is 3 mm. The inner radius of the 2D
axisymmetric winding model is 700 mm and the thickness
Table 4. Test conditions for oil natural (ON) cooling modes of the directed flow washers is 0.5 mm.
Vin (m/s) Re Gr Ri (Gr/Re2) The CFD simulations cover the practical ranges of total
0.017 54 37,173 13.0
oil flow rate and heating power in the winding for typical
OD cooling modes. Similar flow patterns to those shown
0.021 60 28,923 8.0 in section 4.2 were obtained.
0.025 68 20,538 4.4
From dimensional analysis, it can be shown that for a fixed
0.033 87 15,281 2.0 winding geometry with a fixed power loss distribution the
hot-spot factor is controlled by Re and the Prandtl number
The temperature gradient in calculating Ri is set to be the
(Pr) of the liquid. The CFD results have been used to
gradient between the average winding temperature and
determine the variation of the hot-spot factor against Re and
the average oil temperature.
Pr for the investigated OD winding model as shown in Fig.
11 [5]. It can be seen that Re is much more influential than
Pr. When Re ranges from 200 to 600; the hot-spot factor
is approximately constant. When Re is larger than 600, the
hot-spot factor increases monotonically with increasing Re
because of an increasingly uneven flow distribution.

Fig. 9. Temperature and flow distributions in pass 3 from experimental tests.

‘IV’ stands for inlet velocity.

6. Further analysis using CFD

models
Having established the correlation between CFD models
and experimental results using the test rig, more detailed
CFD models have been developed to model winding
Fig. 10. Disc-type winding geometry. (a) 3D geometry of a segment of one
discs composed of paper wrapped strands and using pass between two adjoining sets of spacers. (b) 2D axisymmetric geometry of a
axisymmetric 2D simulations to model a cylindrical 3-pass winding model. (c) Geometric details of the cross section of a strand. [1]

Cigre Science & Engineering • N°13 February 2019

23

Fig. 11. Variation of hot-spot factor with Re and Pr. [5]
6.2 ON Cooling Modes
The LV winding of a 66 MVA 225/26.4 kV ONAN/ONAF
transmission transformer [6] was used as the model for
this study. This winding comprises 4 passes. Pass 1 has
two parts as shown in Fig. 12: part 1 at the bottom has
only 2 discs; part 2 has 19 discs and it is repeated to in the
following 3 passes. Washers are inserted between passes
to produce a zigzag flow pattern.
Fig. 12. Axisymmetric 2D geometry. (a) Winding geometry of pass 1at the
bottom of the LV winding. (b) Geometric details of one strand. [2]
Cigre Science & Engineering • N°13 February 2019
24
Ranges of flow rate and power loss typical to ON cooling
modes were simulated. These simulations were guided
by dimensional analysis which shows that both flow
distribution and temperature distribution for buoyancy
driven flows are controlled mainly by Ri and Re.
Similar flow patterns to those found in the experimentally
tested cases shown in Fig. 9 were found in the ON CFD
simulations. The results show that the hot-spot factor is
predominantly controlled by Ri as shown in Fig. 13 [2]
at least for lower flow rates. The Richardson number
is proportional to the temperature gradient between the
average winding and the average oil and it is inversely
proportional to pass inlet velocity squared.
It can be seen from Fig. 13 that the effect of Re is not
significant until Ri is larger than approximately 0.7. The
variation of the hot-spot factor against Ri can be divided
into three regimes. In regime 1, Ri is close to zero,
representing a condition of low temperature gradient and
high pass inlet velocity, is tending towards OD conditions
rather than ON where the temperature gradient is the
driving force for the oil circulation. In regime 2, Ri is in a
range that indicates balanced buoyancy forces and inertial
forces where relatively uniform flow distributions are
observed resulting in hot-spot factors close to 1 (assuming
uniform heating of the disks). In regime 3, buoyancy
forces dominate the flow with high temperature gradients
and relatively low pass inlet velocity. When Ri is high
enough reverse flow starts to occur and significantly
increased hot-spot factors are observed.
Fig. 13. Variation of hot-spot factor with Re and Ri.[2]
7. Conclusion
The liquid flow distribution in the radial ducts of
transformer disk winding models has been investigated by
experimental measurements and CFD simulations. Uneven
flow distributions at both high and low flow rates including
flow stagnation and reverse flow have been observed and
simulated. A good correlation between observed flows
measured by particle image velocimetry and predicted by
CFD has allowed more detailed CFD studies to explore the
relation between dimensionless parameters and hot-spot
factor. The measurements and simulations were guided by
dimensional analysis, which provides simplicity, insight
and universally applicable results.
For high liquid flow rates typical of OD cooling modes,
the flow distribution and hot-spot factor are found to be
mainly controlled by the Reynolds number (Re) at the
winding pass inlet which is proportional to the total liquid
flow rate and inversely related to liquid temperature. As
Re increases the flow distribution becomes more uneven
with increasing flow fractions at the top part of the
winding pass and reducing flow at the bottom. At high
enough Re for relatively large radial duct to axial duct
dimension ratios, flow stagnation and reverse flows at
the bottom part of the winding pass occur. Under some
circumstances, it has been shown both experimentally and
using CFD that higher total liquid flow rates can increase
rather than reduce hot-spot temperatures. The hot-
spot factor stays relatively constant at low Re and then
increases with increasing Re because of the increasingly
uneven flow distribution. For this reason attempting to
reduce hot-spot temperatures and increase transformer
rating by increasing oil flows for example by increasing
pump power may be counterproductive.
At low liquid flow representative of ON cooling modes,
the dimensionless flow distribution and hot- spot factor are
Cigre Science & Engineering • N°13 February 2019
25
found to be mainly controlled by the Richardson number
(Ri) which is proportional to the temperature gradient
between the average winding and the average oil and
inversely proportional to the total liquid flow rate squared.
When Ri is high (low input flow rates), buoyancy forces
dominate the flow distribution and result in a lower flow
fraction at the top part of the winding pass. If the input oil
flow rate is low enough, stagnation and even reverse flow
can occur at the top of the winding causing an increase in
hot-spot factor and a disproportionate increase in hot-spot
temperature. Under ON conditions therefore Ri needs to
be kept low by ensuring sufficient inlet oil flow to achieve
a reasonable hot-spot factor.
8. Bibliography
[1] X. Zhang, Z. Wang, and Q. Liu, “Prediction of Pressure Drop and
Flow Distribution in Disc- Type Transformer Windings in an OD
Cooling Mode,” IEEE Transactions on Power Delivery, vol. 32, no. 4,
pp. 1655-1664, 2017.
[2] X. Zhang, Z. Wang, Q. Liu, P. Jarman, and M. Negro, “Numerical
investigation of oil flow and temperature distributions for ON
transformer windings,” Applied Thermal Engineering, vol. 130, pp.
1-9, 2018.
[3] X. Zhang, M. Daghrah, Z. Wang, Q. Liu, P. Jarman, and M. Negro,
“Experimental Verification of Dimensional Analysis Results on
Flow Distribution and Pressure Drop for Disc Type Windings in
OD Cooling Modes,” IEEE Transactions on Power Delivery, In press,
2017. (DOI: 10.1109/TPWRD.2017.2739483)
[4] M. Daghrah, Z. Wang, Q. Liu, C. Krause, and P. W. R. Smith,
“Characterization of Oil Flow Rates within Radial Cooling Ducts of
Disc Type Winding Transformers using Particle Image Velocimetry,”
IEEE Electrical Insulation Magazine 2017, Under Review.
[5] X. Zhang, Z. D. Wang, and Q. Liu, “Interpretation of Hot Spot Factor
for Transformers in OD Cooling Modes,” IEEE Transactions on Power
Delivery, In press, 2017. (DOI: 10.1109/TPWRD.2017.2710087)
[6] P. Picher, F. Torriano, M. Chaaban, S. Gravel, C. Rajotte, and B.
Girard, “Optimization of transformer overload using advanced
thermal modelling,” in CIGRE Conference, Paris, France A2-305,
2010.
Optimizing the energizing scheme of
unearthed shunt capacitor banks

R. DOCHE* (RD), R. PATER (RP), S. POIRIER (SP), Y. FILION (YF),

M. LAMARCHE (ML), A. MERCIER (AM)
Hydro-Québec
Canada

Summary The two cases involve a single-break CB switching a

Controlled switching (CS) has been used on Hydro-
Québec’s high voltage (HV) network for more than
twenty years. Initially it was applied for shunt reactors
and later for capacitor banks. More recently, CS for power
transformers was developed. Controlled switching has
been very successful, to a large extent, in replacement
of more traditional mitigation techniques, and has been
generally proven to be quite efficient in controlling inrush
currents and transient overvoltages. CS is well documented
and was studied by several working groups. In the near
future a new IEC standard will be published on alternating
current circuit-breakers (CBs) with intentionally non-
simultaneous pole operation.
On Hydro-Québec’s transmission network, one particular
case was studied following several synchronization alarms
when switching a 120 kV un-earthed capacitor bank with a
single-break SF6 CB[1]. In-field measurements and EMTP
simulations exposed the possible risks with the current
closing scheme in the presence of multiple prestrikes and
CB mechanical scatter. Two possible hazardous situations
are described that could lead to high inrush currents and
overvoltages. An improved closing scheme was proposed
which has all three phases targeting voltage zero.
This paper discusses the challenges one faces when
commissioning a CB for CS. Also field test results are
presented when applying the new energization scheme for
two different study cases involving switching unearthed
capacitor banks on Hydro-Québec’s transmission network.
120 kV unearthed capacitor bank and a double-break CB
switching a 230 kV unearthed capacitor bank. Field tests
are performed with the UNIC system [5] which allows for
precise measurement of making instances and prestrikes
of each phase.
Results confirm that the new operating scheme performs
as expected. Inrush currents are kept below acceptable
levels.
1. Introduction
Due to daily load variation, CB operations must be
performed frequently to control the reactive power
levels. Without mitigation, these operations can result
in important temporary overvoltages which can lead to
undesirable consequences to the equipment and to the
network as a whole. Mitigation techniques must be used,
and new optimal and economical solutions must be studied.
While conventional solutions (e.g. closing resistors)
provide some degree of mitigation, a more affordable
and effective solution is to use CS. Controlled Switching
Systems (CSS) have been successfully deployed and
are now commonly used for many applications such as
capacitor bank, reactor, transformer or line switching.
CSS becomes a challenge especially when retrofitting
existing breakers as CB characteristics such as the Rate
of Decrease of Dielectric Strength (RDDS), mechanical
scatter and effects of temperature, control voltage, idle
time etc. can affect the ability of the system to meet the
target point.

*doche.rene@hydro.qc.ca

KEYWORDS
Controlled switching, rate of decrease of dielectric strength, transient electromagnetic emission, transient switching
overvoltage, unearthed shunt capacitors bank.

Cigre Science & Engineering • N°13 February 2019

26
 Figure 1: Example of temperature compensation for a single break SF6 CB
A recent case analysis study [1] has shown that the closing
scheme, when switching an unearthed capacitor bank,
can be further optimized. The risks associated with the
existing scheme are presented as well as the foreseen
benefits of the new switching scheme. Two study cases
are analysed when applying the new closing scheme. The
data provided by the CS and UNIC systems are used to
perform a detailed analysis.
The diagnostic tool called UNIC, based on transient
electromagnetic emissions (TEE) analysis, allows
for precise measurement of each prestrike instance in
each interrupter. The technology allows for the precise
measurement of each prestrike instance in the first phase
to close even though there is no current flow in the circuit.
The UNIC system has already proved its usefulness being
successfully used for more than 20 study cases on Hydro-
Quebec’s grid [1]-[3].
2. Challenges when
commissioning a cb for
synchronised switching
To reach its goal, a CSS must store and process a lot of
data such CB mechanical and dielectric characteristics,
the application (load circuit) related parameters, internal
settings and must measure in real time the influencing
parameters including control voltage and temperature data.
Figure 1 gives an example of temperature compensation
needed for or a given SF6 CB used for control switching
on HQ’s transmission network. The CSS then calculates
the optimal moment when the close and trip command
is given to the CB coils according to a pre-determined
control strategy for a given application. Additionally
the CSS can store relevant data which can be used for
monitoring and diagnostic analysis.
Cigre Science & Engineering • N°13 February 2019
27
When retrofitting older CBs, mechanical and electrical
CB characteristics are not always available from the
manufacturer, this data can still be obtained during
commissioning tests although this is not the preferred
method because this prolongs the commissioning time.
Optimizing the commissioning time is a critical aspect
that could become a hindrance to the large deployment
and implantation of this technology. At the same time,
these tests are of great importance because the CSS overall
performance often depends upon proper adjustments
made during commissioning. When retrofitting on older
CB, aging can also have an impact on compensation
curves and CB mechanical and electrical characteristics.
Methods and software tools have recently been developed
to reduce the number of tests required at commissioning
before reaching the final settings that provides maximum
system performance.
Typically there are four different types of CS applications:
reactor, capacitor, transformer and line switching.
Different control strategies, commissioning procedures
and examples for each application are included and
explained in technical brochure [to be published] prepared
by CIGRÉ WG A3.35. Regardless of the control strategy
used, obtaining the correct CB compensation curves
and operating characteristics are keys to a successful
implementation of this technology.
Taking as an example shunt capacitor banks application
with isolated neutral, only a few on-line closing tests are
required to evaluate the actual RDDS of the CB. Using
the hypothesis that the mechanical operating time will be
exactly the same for each test, in theory only two on-line
tests are necessary: a closing operation near voltage zero
(where the RDDS influence is at its minimum) and one
other near voltage peak (where the RDDS influence is at
its maximum). Depending on the required precision, more
 Figure 2: Correlation between the mechanical and electrical targets
tests may be required because of the effect for example of
mechanical scatter. Other measurement tool (as the one
described in this paper) will also contribute to enhance
our CB knowledge, thus reducing the number of tests
performed during commissioning.
As illustrated in fig 2, for controlled energization of
unearthed shunt capacitor banks, pre-arcing characteristics
and statistical mechanical scatter of the closing time
need to be considered for determination of electrical
energization instances on the voltage signal (electrical
target).
3. Optimizing the closing scheme
for unearthed capacitor banks
To energize unearthed neutral shunt capacitor banks, it is
common practice to apply the closing scheme described
in Figure 3 as recommended by CIGRÉ WG 13.07 [4].
The first two phases are closed first at zero phase-phase
voltage and then the third phase 90 electrical degrees later
(Fig. 3). Optionally CIGRÉ WG 13.07 also mentions
that “any single phase may be closed first, followed by
closing of the second phase at the instant when the voltage
difference between the first phase and the second is zero”.
Figure 3 Controlled switching scheme at 60 Hz
Cigre Science & Engineering • N°13 February 2019
28
This closing scheme requires that the first two phases close
at the same instant. In practice it is much more difficult to
meet the target point for phase B due to decreasing voltage.
As shown in Fig. 4 closing the first two phases at the same
instant results in a transient overvoltage followed by an
instantaneous voltage rise across the contacts of the last
phase from 1 to 1.5 p.u. (relative to system voltage) due to
the shift in the neutral voltage. This imposes the use of a
CB with a minimal RDDS of 1.5 p.u.
As described in [1], two hazardous situations exist when
using this switching scheme if considering mechanical
scatter.
To avoid these hazardous situations, an improved closing
scheme is proposed (see Figure 5). The closing scheme
is aimed at limiting voltage transients by targeting zero
voltage across each phase of the CB. It is possible to
achieve this by advancing the closing command of phase
C to target its natural zero crossing. Next, phases B and A
would target the same instances as in the original scheme
(Fig. 3). While CIGRE WG 13.07 [7] suggests a random
closing instant of the first phase to close, simulations
rather show that it is better for the first phase to be closed
at zero voltage.
Figure 4 Normal simultaneous closing of B and C,
voltages between the CB contacts
 Figure 5 Proposed scheme of closing: C followed by B and A.

Another advantage is to lower the probability of multiple
prestrikes. In effect, all making instants happen at a
voltage zero crossing and thus the prestrikes instances
occur very close to contact touch. Additionally, the target
window for phase B can now be set on increasing voltage.
In Fig. 6 results of two simulations are shown for a
making instance of phase B 0.8 ms before and after the
target point. For the case on the left, there is no risk of
premature making of phase A. For the case on the right
voltage is headed for 1.73 p.u. but voltage across the
interrupter of phase A is reduced sharply at the making
instant of phase B.
The proposed scheme has the advantage of considerably
reducing transient overvoltages allowing for a more
optimal setting of the target window especially on the
second phase to close. Also the target can be set on
increasing voltage. Both of these advantages should result
in better overall control and performance.

Figure 6 Proposed closing scheme with B phase missing the target point by ±0.8 ms

Cigre Science & Engineering • N°13 February 2019

29
4. Optimized scheme study case
analysis
The new proposed controlled switching scheme has
been applied in the field to two CBs switching unearthed
capacitor banks. The first case (case 1), is an existing 145
kV single break SF6 CB switching a single capacitor bank
of 96-Mvar at 120 kV. The second (case 2), is a newly
installed 362 kV, two-break, SF6 CB switching a 66-Mvar
capacitor bank at 230 kV. This capacitor bank is connected
back-to-back to another 66-Mvar capacitor bank. In both
cases, a series current limiting reactor is present.
The analysis is based on two types of complementary data:
recordings supplied from the CSS itself and that from
the UNIC diagnostic system which is based on transient
electromagnetic emissions (TEE). The former provides
the current and voltage wave forms and the latter line
and coil current as well as the signal indicating the exact
moment of prestrike in each CB interrupter. However,
due to logistic constraints the CSS recordings were not
available for all CB operations.
Regarding the 362 kV CB (Case 2), three operations
are recorded showing good performance of the scheme.
However the details are not discussed here. Recorded
data for the 145-kV CB consists of 14 closing operations
with the initial control scheme, 7 operations during
commissioning of the new control scheme and a total of
Figure 7 Voltage current waveforms for upper limit alarm situation
Cigre Science & Engineering • N°13 February 2019
30
25 operations with the new control scheme over a period
of two months.
a) Case 1 - Test results for 14 operations with the
original closing scheme
Results show that the making instance of B phase is
subjected to multiple prestrikes in 8 out of 14 operations.
The delays between the prestrikes vary between 0.6 ms
and 1.35 ms. The last prestrike is followed immediately by
current initiation in C phase. For the 6 closing operation
without multiple prestrikes (on B phase), the making
instance of C phase precedes that of B phase. Such a
situation is described in [1] as a potentially hazardous
situation leading to a premature making instant of the last
phase to close.
b) Case 1 - Commissioning test to implement the new
closing scheme
The new closing scheme implies changes to the target point
of the first phase to close (phase C) which now targets
voltage zero. This timing adjustment cannot be done
using V/I waveforms because there is no current initiated
after the making instant with an unearthed system. Single
or multiple prestrikes are visible and measured which
gives a good indication if the making instant of C phase is
within the target window.
During commissioning tests, the upper limit on B phase
was checked by gradually delaying the closing command
of phase B from its original target. Target points of phase
 Figure 8 TEE recording representing prestrike delay between A and B

A has remained unchanged. C phase was eventually
adjusted in order to better target voltage zero. When the
delay between phase B and A reached a value of 1.4 ms, an
alarm occurred in the controlled switching device (CSD).
As illustrated in Figure 7, all three phase currents started
simultaneously one millisecond after voltage zero crossing
of phase B. Thus, the prestrike in phase A occurred 3 ms
ahead of the expected target. In Fig. 7, currents recorded
by the CSD show a current peak reaching up to 5 p.u. in
phases A and C. The inrush current magnitude measured
by the CSD is generally underestimated due to its low
sampling rate. TEE measurements (Fig. 8) show that
phase A prestriked 2.1 microseconds before phase C, thus
making instantly an inrush current with phase C, which
already had closed contacts at that time.
In the simulation reproducing the upper limit in Fig. 9,
the prestrike instance is delayed by intervals of 0.2 ms up
to 1 ms after the target point. This simulation shows the
effect of this delay on the voltage appearing across the CB
terminals of phase A. It is shown that if phase B prestrikes
around 1 ms after its target point, phase A is at risk of
premature making, thus causing a strong inrush current
reaching up to 9 p.u. according to the simulation of Fig.
10. Delaying the target point of the last phase to close, by
one half cycle, would eliminate this risk.

Figure 9 Simulation for upper limit alarm situation

Figure 10 Simulation of possible inrush current

Cigre Science & Engineering • N°13 February 2019

31
 Figure 11 Electrical making instances relative to temperature of operation

c) Case 1- 25 operations performed over a two month
period with the new control scheme (all three phases
targeting zero voltage)
The 25 closing operations confirm that the new scheme
is stable and the CB performs as expected. The slope
of the making instances vs ambient temperature for the
three phases has a similar slope to that of the temperature
compensation curve programmed in the controller
as shown in Fig. 11. Differences in making instances
between each phase depends on two factors, true
mechanical operating time and pre- arcing time. Phase
A is well adjusted as by voltage waveforms. The closing
command of phase C could possibly be advanced by 0.2
ms. The voltage waveforms show that phase B prestrikes
at around 0.32 p.u. which is about 0.2 ms after the target
of 0.5 p.u. 58
As the CB is not equipped with a displacement transducer,
it is not possible to directly calculate the pre-arcing time
for each operation. For A and B phases the making instant
can be obtained from the CSS by measuring the instance
of current initiation.
Although results were as expected for 24 of 25 operations
one particular operation resulted in a making instant of
B phase at a “lower limit”. As it is described in section
3, there is a probability that the second phase to close
prestrikes prematurely if its RDDS become too close to
the decreasing voltage across the CB. As depicted by
TEE measurements shown in Fig. 12, phase B prestrikes
prematurely, followed by phase C 2.2 microseconds later.
The last prestrike in phase A was not influenced by the
early prestrike of phase B.
The lower limit case has been reproduced by simulation.
In Fig.13, the first prestrike in phase B causes a fast rising
voltage across the CB terminals of phase C. Therefore,
the inrush current between phases B and C is initiated
rapidly and has a peak value of around 4.5 p.u. (Fig.
14). However, the transient voltage across phase A at the
moment of the first prestrike in phase B is much lower
than its voltage withstand (considering the typical RDDS
of 1.5 p.u. shown in Fig. 14). A premature prestrike in
phase A is thus unlikely to happen.

Figure 12 TEE recordings representing prestrike delay between B and C

Cigre Science & Engineering • N°13 February 2019

32
 Figure 13 Voltages across the contacts for the lower limit situation

Figure 14 Inrush currents for the lower limit situation

5. Conclusion
Results are very promising when applying a new
switching scheme for unearthed capacitor banks. Two
case studies were performed, one on a 145 kV single-
break CB and another on double-break 362 kV CB. Even
though one operation came close to the theoretical limits,
inrush currents were kept at acceptable levels during the
trial period.
The analysis also revealed that breaker commissioning is
critical especially when the CBs are not equipped with
transducers. Aside from the data available via the CSS
such as line voltage and current, software tools and the
use of special acquisition equipment such as the UNIC
system will help to reduce commissioning time.
6. Bibliography
[1] R. Doche, R. Pater, S. Poirier, M. Lamarche, Y. Filion “Transient
Electromagnetic Emission Analysis Used to Study Controlled
Switching of Unearthed Shunt Capacitor Banks”, 2018, IEEE PES
Transmission & Distribution Conference & Exposition, Denver
[2] A. Mercier, S. De Carufel, Y. Filion, E. Portales, J. Caron, P. Taillefer,
“Upgrading HVAC circuit breakers using controlled switching:
Hydro-Québec’s 20 year experience”, 2014, IEEE PES Transmission
& Distribution Conference & Exposition, Chicago, USA
[3] R. Doche, R. Pater, S. Poirier, M. Germain, S. Gingras, “Field
Experience with Non-intrusive In-service Diagnostics of Circuit
Breaker Based on Transient Electromagnetic Emissions”, 2017, 3rd
International Conference on Condition Monitoring, Diagnosis and
Maintenance, Bucharest, Romania
[4] “Controlled Switching of HVAC Circuit Breakers Guide For Application
Lines, Reactors, Capacitors, Transformers (1st Part)”, CIGRE WG
13.07, Électra No. 183, April 1999
[5] S. Poirier, R. Pater, R. Doche, “A Non-Intrusive Diagnostic Approach
to Prevent Circuit- Breaker Critical Failures”, 2010, International
Conference on Condition Monitoring and Diagnosis, Tokyo, Japan

Cigre Science & Engineering • N°13 February 2019

33
Current and future applications of HPTE
insulated cables systems
A. BAREGGI*, P. BOFFI*, S. CHINOSI**, S. FRANCHI BONONI*, L. GUIZZO***, G. LAVECCHIA***,
M. MARZINOTTO***, G. MAZZANTI****, G. POZZATI*
*Prysmian SPA,
**Prysmian Cavi e Sistemi Italia SRL,
***Terna rete Italia SPA,
****University of Bologna,
Italy
Summary
HPTE insulated cables represent an attractive alternative
to traditional cables insulated with crosslinkable materials.
Thanks to a set of advantages, mainly due to the use of
thermoplastic PP based insulation, superior performances
can be achieved for cables with HPTE insulation. After
an important experience in the field of MV cables, now
HPTE insulation is being extended to HVAC and HVDC
areas.
In the field of HVAC, a cable length (1000 mm2 Cu 150
kV with HPTE insulation) has been installed in the Italian
Transmission network and is in service since 2013.
In HVDC important qualifications have been achieved
in 320 kV, 350 kV and recently 600 kV voltage class, at
high conductor temperature (90°C) and including both
VSC and LCC typology. These results have been obtained
thanks to an extensive work performed on materials
development and technological improvements, for cables
and accessories.
The paper describes the developments performed and the
results achieved, that are very promising especially in
view of possible applications of HPTE insulated cables
for EHVDC systems.
1. Introduction on HPTE insulation
technology
The opportunity to develop an insulation technology for
MV and HV cables alternative to the well- known and
*alberto.bareggi@prysmiangroup.com
KEYWORDS
HV cable - HVAC - HVDC - Thermoplastic - Insulation – HPTE
Cigre Science & Engineering • N°13 February 2019
34
experienced materials like XLPE and HEPR has driven
cable manufacturers to investigate in the early 2000s the
use of a polyolephin (PP) considered so far not adequate
to provide the requested properties of a high performance
electrical insulation for cables. Main challenges to
overcome in order to demonstrate the feasibility of use
of PP based insulation for electrical cables were the
following :
- Dielectric properties on cables
- Resistance to electrical ageing
- Resistance to thermo-oxidative ageing
- High stiffness of PP based materials
- Working temperature not adequate in the field of low
temperature, due to high fragile temperature of PP.
- Melting temperature of PP materials lower than max
short-circuit allowable temperature (250 °C)
- PP morphology and characteristics during the
insulation process (extrusion and cooling)
Through the selection of a special polymer blend and
the application of an advanced production process, it
was possible to overcome the here above mentioned
challenges and to finalize the development of the new
insulation type, called HPTE, firstly introduced in the MV
cables area.
The development of HPTE was based on the selection of
polypropylene raw materials that were formulated and
compounded in view of application to power cables. Since
the excellent electrical properties of polypropylene films
are well known, a main goal of the early R&D activity was
focused on achieving an adequate material morphology,
in consideration of the insulating layer thickness typically
used for energy cables.
The selected formulations turned polypropylene into a
versatile and high performance insulating material for
power cables, both in terms of physic-chemical properties
and electrical characteristics. Indeed, the new insulating
compound is characterized by good flexibility and very
high values of tensile strength and elongation at break
together with high thermo-pressure properties.
It must be underlined that, in general, thermoplastic
polyolefin lack of these properties all together. Low density
polyethylene, for example, needs to be crosslinked to get
cables with satisfying thermo-mechanical properties at
90°C as operating temperature, and at 130°C, as overload
condition.
From the process perspective, the use of thermoplastic
insulation in place of crosslinked versions allows to
overcome a series of drawbacks, such as the creation of
pre-scorches or ambers in the extrusion apparatus.
Another process advantage of HPTE insulation is related
to material filtration. While there are practical limits in
filtering crosslinkable polyethylene, due to the possible
scorch phenomena, HPTE insulating compound can
be filtered much more thoroughly, with advantages in
terms of material cleanliness. More efficient filtration
is an important advantage, because the presence of
contaminating particles in cable insulation, severely
affects electrical performances and, in general, cable
reliability.
A further and important advantage of HPTE material and
processing is that no degassing treatment is required since
no crosslinking step is carried out. This zero degassing
technology, at the basis of this new cable concept, permits
the manufacturing of power cables without the need of
extensive and long lasting thermal treatments for the
evacuation of crosslinking by-products..
Full recyclability of polymeric materials is also a valuable
property of this technology, with benefits in terms of
carbon foot print. Besides recovering metals such as in
other energy cables, all the polymeric layers of these
cables, including insulation, can be fully recycled, i.e.
used for production of components like fillers applicable
to various frameworks.
Cigre Science & Engineering • N°13 February 2019
35
Finally, in terms of product properties, the new insulation
allows to achieve superior thermo- mechanical properties
for the cable at temperature significantly higher than
other crosslinked materials; the high melting temperature
of HPTE insulation opens the possibility to increase the
actual temperature limitations of energy cables (70 –
90 °C) to a further step; this represents an undisputable
advantage in the network, either in terms of initial
investment or in terms of safety margin of the power
transmission / distribution system.
Thanks to the here above listed properties, HPTE
insulation has been successfully introduced in MV field
in Italy and The Netherlands, and is under discussion in
other countries. In particular, in Italy the first 20 kV cable
with HPTE insulation was installed in 2006 and is in
service since then; so far, about 36000 km (measured like
single core) of cables have been installed in Italy.
HPTE technology was successfully introduced also in the
Netherlands for MV cables, based on local specification;
this includes accelerated ageing test in water, that was
overcome by HPTE insulated cables with important
margin, demonstrating the natural and interesting attitude
of the PP based insulation to resist to water treeing
phenomena.
Based on the above mentioned properties and
achievements, since HPTE is not just a formulation but a
technological platform for power cables, this innovative
category of material has been taken in consideration for
several new applications.
By properly tuning the physic-chemical properties of
the ingredients, it was possible to develop and optimize
specific HPTE versions for insulation of HVAC or HVDC
power cables, without constrains related to residual by-
products. The extremely low level of polar additives
makes the formulation especially suitable as insulation of
HVDC cables.
2. 150 kV experience and
feedback from the field
In 2013 an experimental portion of HVAC cable with
HPTE insulation was installed in 132 kV link named
“Lacchiarella – Vernate T.167”, located in north Italy
near Milan [1]. A section of about 500 m length has been
inserted between two joint bays as a portion of total length
 Figure 1 – Lacchiarella circuit straight line diagram

of 5080 m. The cable is an 87/150 kV, 1000 mm2 copper unit has been provided. The sensors have been installed on
conductor, HPTE insulation, welded aluminum sheath the cable surface at the transition joints location [3]. These
and PE outer sheath while rest of the link is composed by collect and transfer all PD discharge activity signals to the
1600 mm2 aluminum conductor, XLPE insulation, copper acquisition unit located inside the link box manholes. This
wire and aluminum composite screen and PE outer sheath. last will send wireless all the data to a remote location.
The junction between the two cable designs (HPTE and The electrical feeding of the unit is guaranteed by a
XLPE) has been obtained by the use of asymmetric and special device installed around one of the three power
bi-metallic sectionalized cable joints of “click-fit” type. cables, generating the necessary cuWrrent value (fig.2).
The circuit is installed mainly in trefoil formation, directly
buried in trench with some road crossings and HDD in
plastic pipes and short portions in air at the terminations.
From the electrical viewpoint, the system has been
divided into 10 sections (9 sections for cross bonding and
1 section for single point bonding). Hereunder figure 1
shows the single line diagram:
Before to put in service the link, an intensive test program
was conceived on complete cable system in order to assess
the quality of this innovative insulation; in absence of a
specific standard dedicated to HV cables rated at 90°C
and insulated with thermoplastic materials, it was decided
to carry out the test in accordance to IEC 60840 [2] and
to national TSO specifications. Type test, prequalification
test procedure according IEC 60840 plus additional
thermal cycles according to table 1 have been positively
completed.
In order to monitor continuously the partial discharge
activity, a permanent system, composed of wireless PD
sensors, an electrical feeding device and an acquisition Figure 2 - PD acquisition system

Table 1 – Sequence of tests performed on 150 kV cable with HPTE insulation

Test Results

20 therm. cycles (95°C – 100°C; 174 kV) No breakdown or flashover

180 therm. cycles (95°C – 100°C; 150 kV) No breakdown or flashover
20 therm. cycles (100°C – 105°C; 174 kV) No breakdown or flashover
20 therm. cycles (110°C – 115°C; 174 kV) No breakdown or flashover
20 therm. cycles (110°C – 115°C; 217 kV) No breakdown or flashover

Cigre Science & Engineering • N°13 February 2019

36
 Figure 3 - PD acquisition pattern
At specified time intervals the system acquires the PD
measurements from the six transition joints located on the
three phases of the line and the measures are uploaded on
a cloud system (fig.3).
The link has been running continuously since 2013 and
even though the line was not particularly loaded so far
(about 35% of nominal current – median value), feedback
from field is positive and no issues or PD anomalies were
noticed during this period; the good performance was
confirmed also by positive results of aforementioned test
program.
3. Characteristics of HVDC
extruded insulation
In recent years, the relatively-new HVDC extruded
(polymeric) cable systems have become more and more
competitive compared to the traditional mass impregnated
non-draining (MIND) HVDC cable systems. Indeed,
although the latter have totaled much more service
experience, the former have some major advantages,
e.g.: they are much more environmentally-friendly as no
impregnating fluid is used in the insulation; the maximum
permissible conductor temperature in normal operation is
higher; splicing is much easier [4].
For these reasons, extensive and sound R&D activities
have brought now the voltage and power ratings of
installed HVDC extruded cable systems to the 320 kV
and 1000 MW levels, and these levels are expected to
increase. This is shown by the 320-kV XLPE-insulated
HVDC cable system projects named “Sylwin 1”, “Helwin
2”, “Dolwin 1”, “Dolwin 2”, and “Dolwin 3” linking
wind farms in the German North Sea to the mainland and
by the 320 kV/ 2000 MW XLPE double bipolar HVDC
cable link named “INELFE” between France and Spain.
In addition, the “Nemo Link”, a 130 km subsea and 11.5
km land 400 kV/1000 MW XLPE HVDC cable system, is
under construction between UK and Belgium [5].
For an improved design of HVDC extruded cable systems
capable of achieving these goals, R&D has followed
mainly three routes [4] :
Cigre Science & Engineering • N°13 February 2019
37
1) as for the converter, improvements in HVDC VSC
converter station technology have removed the
requirement for polarity reversal on DC systems up to
1000 MW (and expected to increase);
2) as for the cable system, a particular care has been
devoted to the accessories (joints and terminations)
that are often the weakest points of the whole cable
line;
3) as for the cable insulation, the performances of DC-
XLPE have been dramatically improved with respect
to traditional compounds used for AC.
Focusing on this latter aspect, among the various methods
followed starting from the early 1980s to develop HVDC
cables with polymeric insulation, early research trialed the
use of LDPE, XLPE or HDPE without any modification.
The poor results obtained led to abandon this strategy
and select the so called “DC-XLPE”, namely XLPE
compounds specifically developed for DC application,
first of all by modifying electrical and thermal resistivity of
the polymer so as to limit the field inversion phenomenon
(the so-called stress inversion free cable) [6]. In this
framework, space charge build-up in XLPE under DC
stress resulted a major concern for its use in HVDC cables
[6], [7]. As well known, space charges modify the electric
field distribution in the insulation and may lead to local
field enhancements that put the long-term reliability of the
cable system at risk, especially in the presence of voltage
polarity reversals [4] .
Therefore, many different methods for investigating space
charge in solid polymeric materials, set up over the last
decades [9], [10] have given a fundamental contribution to
the realization of present HVDC extruded cable systems.
Inherent difficulties - mainly related to thick insulation -
make measuring space charge on full-size cables difficult,
but recently space charge measurements on full- size
HVDC extruded cable loops have been - and are being -
used worldwide in qualification tests for HVDC extruded
cable link projects. Until 2017 these measurements
were based on project-dependent agreements between
cable manufacturers and the customers, as an agreed
standard procedure was missing. Aiming to fill this gap,
in 2015 a protocol for space charge measurement on
full-size HVDC extruded cables up to 550 kV during
load cycle qualification tests was proposed by the IEEE
DEIS HVDC Cable Systems Technical Committee in a
dedicated position paper [11]. This protocol was improved
and established as a reference IEEE best practice in
2017, when IEEE Standard 1732-2017 was issued. This
Standard enables the use of either the Pulse-Electro-
Acoustic (PEA) or the Thermal Step Method (TSM) for
measuring space charge [12].
To date, space charge behaviour, electrical resistivity
and electrical breakdown stress are deemed as the most
important parameters for designing and manufacturing
HVDC polymeric cables. Therefore current R&D
procedures on HVDC extruded cables are split into three
subsequent stages, as follows [13] :
1) “screening” research tests aiming at the selection of 1 or
2 materials/technologies among a wide range, having
limited space charge accumulation and good thermal
stability as the main performance indicators. Thus,
space charges, electrical resistivity (vs. temperature
and electric field), impulse voltage performance under
superimposed HVDC are investigated on small size
insulating samples representative of the cable system,
i.e.: thin films and plaques (insulation thickness
100 µm  ÷ few mm); cable model “A” (conductor
cross-section ≈ few mm2, insulation thickness
< 2  mm); cable model “B” (conductor cross-section
≈ 50  mm2, insulation thickness ≈ 4÷5 mm);
2) development tests on a full size cable as a validation of
the previous choice, i.e. thermal stability tests, space
charge measurements, heating cycles under HVDC,
polarity reversal test (if required for the application),
superimposed impulse tests. Heating cycles, polarity
reversal and superimposed impulse tests should be
performed up to breakdown. During heating cycles the
space charge behaviour can be monitored;
3) qualification tests, carried out according to CIGRÉ TB
496 [14] or IEC Standard 62895 [15] depending on
voltage level and land/submarine application.
For all these reasons, since the early 1990s, R&D on
HVDC extruded cable systems aimed at the suppression
of space charge build-up in DC-XLPE. As space charges
can be either electronic or ionic, and can be either
injected from the semiconducting shields or created
Cigre Science & Engineering • N°13 February 2019
38
by cross-linking by-products (i.e. residues of peroxide
dissociation) under the action of the electric field, space
charge suppression in the polymer is achieved typically
by means of the following strategies [4], [7], [8] :
1) modification of the features of the semicon/
insulation interfaces so as to reduce charge injection
from semicons, e.g. by: changing the composition
of semicons themselves; changing the electrical
properties of XLPE surface so as to control space
charge behaviour and injection from semicons;
2) reduction of the peroxide content in DC-XLPE
crosslinking;
3) stronger and longer degassing treatments after
crosslinking for DC-XLPE vs. AC-XLPE cables;
4) modification of the bulk XLPE compounds, e.g. by:
blending PE with another polymer; using additives or
(nano)-fillers in PE-based compounds so as to reduce
space-charge accumulation, mitigate by-product
effects and make electrical resistivity more uniform.
Nano-fillers in particular may be either organic or
inorganic, and either conductive or polarized [16],
[17].
Nevertheless, these strategies might not be fully effective
as the reduction of the peroxide content involves that
thermo-mechanical performance has to be reduced
and compatibility with accessories has to be proven in
the long term. Moreover, the introduction of additives
mitigating the by-product effects requires a uniform
dispersion of these additives in the compound, as well
as a proven long-term stability and effectiveness of
such additives. The same holds for nano-fillers, whose
uniform dispersion in a large amount of insulating
material requires a very reliable technology yet to be
fully- established [8].
Hence space charges are still a major issue for cross-
linked polymeric insulation for HVDC cable application
to date, especially in view of the higher and higher
level of electric stress demanded by future HVDC cable
system projects, having a voltage level of 525 kV as the
closest target. Hence, a non-cross-linked thermoplastic
insulation capable of reducing space charge storage
dramatically, thanks to the elimination of peroxide
decomposition, can be greatly effective as for the long-
term reliability of HVDC extruded cable systems.
 Figure 4 : “24 hours” load cycles at – 592 kV and at + 592 kV

4. TT @ 320 kV LCC & VSC on
HPTE cable system, 90°C
After successfully completion of the Prequalification test
performed according to TB Cigré 496 [14] for Voltage
Source Converter (VSC) [8], a test circuit, composed by
100 m of HPTE 320 kV DC cable 1000 mm2 Cu conductor,
1 premolded joint and 2 composite termination, has
been submitted to Type Test sequences in order to cover
requirements for both LCC and VSC systems.
4.1 Summary of the load cycle
Here below the different phases of the load cycle test are
summarized :
- Twelve “24 hours” load cycles at negative polarity
at UT (- 592 kV)
- Twelve “24 hours” load cycles at positive polarity
at UT (+ 592 kV)
- Eight “24 hours” load cycles with polarity reversal
at UTP1(± 464 kV)
- Three “48 hours” load cycles at positive polarity at
UT (+ 592 kV)
A rest period of 24 without voltage but with heating
applied has been done between the two blocks of twelve
“24 hours” load cycles at – 592 kV and + 592 kV. During
the rest period, the heating was applied as thermal cycle.
In Figure 4 the 24 heating cycles and relevant voltage
applied are reported
4.2 Superimposed switching impulse voltage test
After the load cycles previously described, the circuit has
been submitted to the superimposed impulses tests.

Figure 5 : Scheme of the test circuit realized for the superimposed impulse tests

Cigre Science & Engineering • N°13 February 2019

39
 Table 2 : sequence of superimposed impulses
To perform the impulse superimposed on DC test it was
necessary to electrically separate the HVDC generator
from the impulse generator and vice versa (fig.5); in order
to satisfy this condition the test circuit was realized with
a sphere gap.
The superimposed switching and lightning impulse tests
were performed on the whole test assembly. As described
in CIGRE 496 (paragraph 1.5.5.) [14] the cable has been
heated and the conductor temperature has been kept steady
for the last 10 hours before the superimposed impulses at
the specified temperature (≥ 90°C).
In table 2 the sequence of superimposed impulses applied
is summarized :
At the successful completion of the impulse testing the
test assembly has been subjected to 2 hours at a negative
DC voltage of UT (592 kV) with no heating applied. Said
test has been passed as neither breakdown nor flashover
have occurred.
5. Ageing test procedure under
sustained polarity reversal
A new test protocol was developed for DC extruded cables
(including also new thermoplastic materials) with the aim
to compare the results of sustained polarity reversal test
with the current knowledge on MIND cables. The main
interest was a first approach for the comparison of DC
extruded cables and MIND cables over polarity reversal
endurance. Specific numerical models for the evaluation
of life expectations of both cable types have been used
to select both accelerated test duration and the number of
polarity reversal stress to be performed. The test program
is made of two main sections: the ageing section and the
sustained polarity reversal stress section (formally a second
ageing section). The first ageing section is mainly based on
the procedure suggested in [14] for cables to be installed
in systems that allow the voltage polarity reversal, with an
additional acceleration factor around 3 so as to reduce test
times with respect to the standard duration of 1 year. As
far as the second section is concerned, a proper sustained
Cigre Science & Engineering • N°13 February 2019
40
polarity reversal stress procedure has been adopted
formally based on the experience gained by main Italian
TSO through the years on tests made on MIND cables [18].
For HPTE insulation the test has been performed at a
reference value U0=350 kV DC. The tested cable (1000
mm2 copper conductor, HPTE insulation 16,5 mm
thickness, welded aluminium sheath and PE outer sheath)
was assembled with 1 premoulded joint and 2 composite
terminations. The test was completed successfully: neither
breakdown nor evidence of thermal instability happened
during the entire tests on the tested objects.
6. Technological progress for
superior voltage cable prototypes
(600 kV)
Further improvements were added to the original HPTE
technology in order to comply with higher performances
requested to HV AC and DC applications
- Material handling system, with the use of
pressurized and dry air with extreme high level of
filtration;
- Very high level of compounds filtration during
extrusion;
- A new concept for injection system for liquid
additives;
- A new concept of mixing screw for on-line
compounding, developed by using the most recent
3D thermal and fluid dynamics FEM analysis
methods;
- Gradient cooling process after extrusion in
pressurized silicon oil;
Two new 525 kV DC prototypes were positively produced,
with 2000 mm² copper and 3500 mm² aluminium
conductors.
7. Technological progress for
accessories
The development of joints and terminations for
525/600  kV HVDC extruded cables has required the The DC voltage applied on models was calculated in
introduction of specific tests on materials and new design such a way to obtain along the interface and inside the
approaches to improve the reliability of the accessories accessories similar electric stresses present on full-size
under DC electrical field. In fact, their behaviour is cable system during Type Testing according to CIGRE
strictly depending on temperature- related insulation TB 496 [14] requirements. In particular, based on FEM
characteristics and also strongly influenced by the calculations, the value of
phenomenon of the space charges accumulation along the
210kV can be considered as a reference for 525kV voltage
interface between cable and accessory.
class and 230kV the same for 600kV level. Model2 joints
Accessories with EPR pre-moulded components have and termination that showed the best combination of
shown very good performances on 320 kV DC extruded design and materials successfully passed all the required
cables and have been developed up to 525/600 kV voltage steps, including lightning impulse test up to 300kV,
DC thanks to new materials and innovative design and were conveniently scaled for full-size applications.
comprehensive of both geometrical and non-linear electric
Following accessories (fig.6) have therefore been
stress control systems.
developed and successfully tested on 2000mm2 HPTE
In order to verify new joint and termination design and insulated cable with 25 mm insulation thickness:
new materials for HVDC accessories up to 525/600 kV
- Outdoor terminations with composite insulator,
DC, investigation tests have been carried out on reduced
both normal type and explosion-proof version.
size models. HPTE insulated model cable was used, with
- Straight Joints (both un-sectionalized and
5.3 mm nominal insulation thickness.
sectionalized).
Scaled cable and accessories have been submitted to an
The internal tests, carried out with the cable conductor at
internal thermal stability test structured as follows:
90 °C and a thermal drop of 30°C, included:
- Three thermal cycles (4h heating/4h cooling) with
- N° 12 “24 hours” load cycles at negative polarity at
90 °C on the conductor and 55 °C on the outer
971 kV
semiconductive sheet, with voltage applied in
- N° 12 “24 hours” load cycles at positive polarity at
positive polarity.
971 kV
- 8 hours in temperature with the cable grounded.
- N° 8 “24 hours” load cycles with polarity reversal
- Three thermal cycles (4h heating/4h cooling) with
cycles at 761 kV
90 °C on the conductor and 55 °C on the outer
- N° 3 “48 hours” load cycles at positive polarity at
semiconductive sheet, with voltage applied in
971 kV
negative polarity.
The voltage levels according to table 3 have been applied
during the aforementioned heating cycles :
Table 3 : voltage levels applied during execution of heating cycles

Voltage
1° Voltage Step ±190 kV
2° Voltage Step ±210kV
3° Voltage Step ±230kV
4° Voltage Step ±250kV
5° Voltage Step ±270kV
6° Voltage Step ±290kV
7° Voltage Step ±310kV Figure 6 : 525 kV sectionalised joint and 525 kV Outdoor Termination

Cigre Science & Engineering • N°13 February 2019

41
8. Qualification of 600 kV system
In order to qualify the new HPTE material for application
up to 600 kV HVDC it has been performed a complete
Type test sequence extrapolating the test parameters from
CIGRE TB 496 [14] for VSC system which, as known,
is limited up to 500 kV voltage class. The Type Test was
performed on a circuit composed by :
• 525 kV, HPTE, HVDC, 1 x 2000 mm2 CU 600 kV
cable;
• two outdoor sealing ends with composite insulator:
TES 600 kV.
During the type test, it was adopted the load cycle test
for VSC (Voltage Source Converters) systems at a max
temperature on the conductor of 90°C.
In order to circulate an AC current through the conductor
of the cables, the terminations were connected together
(series circuit) and heating transformers were placed on
the cables. The total length of the circuit (including the
terminations) was 47 m, with an active length of cables
of 36 m.
In order to define the heating parameters (current and
Figure 7 : cable system.
Cigre Science & Engineering • N°13 February 2019
42
power to be applied to the heating tapes) able to satisfy the
request of TB 496 [14] for heating cycles, a dummy loop
was realized with a 6 m length of the same cable used for
the electrical type test; the sample was prepared exactly
as the main loop.
This dummy loop was used not only in order to define
the parameters, but also during the tests in order to check
continuously the thermal condition of the main loop.
Figure 7 shows the cable system under test.
8.1 summary of the load cycles
In accordance to extrapolation of CIGRE TB 496 [14] the
following heating cycles sequence has been defined:
- Twelve “24 hours” load cycles at negative polarity
at UT (- 1110 kV)
- Twelve “24 hours” load cycles at positive polarity
at UT (+ 1110 kV)
- Three “48 hours” load cycles at positive polarity at
UT (+ 1110 kV)
A rest period of 24 hours without voltage but with heating
was adopted between the block at negative polarity and
the block at positive polarity.
 Table 4 : Impulses sequence
8.2 Superimposed impulse voltage test
After the load cycle test the circuit was submitted to the
superimposed impulses test.
In order to optimize the testing activity the impulses
sequence has been conducted as in Table 4, where also
parameter and results are shown.
At the positive completion of the impulse testing the test
assembly has been subjected to 2 hours at a negative DC
voltage of UT (= 1110 kV) with no heating applied. Said
test has been passed as neither breakdown nor flashover
have occurred.
9. Conclusions
The big demand for transmission of high electrical power
in long distances has fostered the fast and successful
development in the recent years of HVDC Transmission
Systems at increasing current and voltage levels; in fact,
both traditional and newer technologies are evolving.
In the case of the new technologies, there is a quite
consistent difference between the state of the art in terms
of qualification and in terms of return of experience of
“in service” applications. In other words, the way of the
development of HVDC cables system is in continuous
and pressing evolution. As matter of fact, the technology
for the near future in HVDC cables systems seems to be
solid insulation. In the field of solid insulation, between
the actual technological development axes pursued by
main Cables Manufacturers and Compounds Suppliers,
the HPTE technology is becoming an operational reality.
The recent development demonstrated the feasibility of
600 kV DC cables systems made with HPTE
technology, according to VSC test protocol defined by
Cigré TB496 [14]. This technology permits to produce
a HVDC thermoplastic cable with lower environmental
impact. In addition, the HPTE technology doesn’t require
chemical reactions to achieve the properties required for
long term electrical integrity of HVDC insulation systems.
This feature gives the not negligible benefit to avoid by-
Cigre Science & Engineering • N°13 February 2019
43
products, making this the simplest and most effective
solution against the well-known problem of space charge
traps created by the by-products themselves.
This breakthrough innovation in cable technology will
enable a transmissible power by HVDC systems up to
more than 3 GW (depending on thermal environmental
and installation conditions), which is more than twice
the value achieved with 320 kV DC systems currently in
service.
10. Bibliography
[1] M. Albertini et alii - Development and qualification of 150 kV cable
produced with highly innovative P-Laser technology (CIGRE’ 2014,
Paris) [2] IEC 60840 - Power cables with extruded insulation and
their accessories for rated voltages above 30 kV (Um = 36 kV) up to
150 kV (Um = 170 kV) – Test methods and requirements - Ed.4.0
2011-11.
[3] R. Candela et alii - A novel partial discharge detection system based
on wireless technology (AEIT Annual Conference 2013)
[4] G. Mazzanti, M. Marzinotto - Extruded Cables for High
Voltage Direct Current Transmission: Advances in Research and
Development, Power Engineering Series - Wiley-IEEE Press, 2013,
ISBN 978-1-118-09666-6.
[5] G. Mazzanti – Editorial, Special Issue on Worshop on HVDC cables
and accessories (IEEE Electr.
Insul. Mag., Vol. 33, No. 4, pp. 4-5, Jul./Aug. 2017).
[6] M. Salah Khalil - International research and development trends and
problems of HVDC cables with polymeric insulation (IEEE Electrical
Insulation Magazine, Vol. 13, No. 6, pp. 35-47, November-December
1997).
[7] T.L. Hanley, R.P. Burford, R.J. Fleming, K.W. Barber - A general
review of polymeric insulation for use in HVDC cables (IEEE
Electrical Insulation Magazine, Vol. 19, No. 1, pp. 14-24, January/
February 2003).
[8] M. Albertini, A. Bareggi, L. Caimi, L. De Rai, A. Dumont, S. Franchi
Bononi, G. Pozzati, P. Boffi - Development and high temperature
qualification of innovative 320 kV DC cable with superiorly stable
insulation system (9th International Conference on Insulated Power
Cables Jicable’15, paper No. A7.3, pp. 1-6, Versailles, France, 21-25
Jun. 2015).
[9] T. Takada - Acoustic and optical methods for measuring electric
charge distributions in dielectrics (IEEE Trans. Dielectr. Electr. Insul.,
Vol. 6, No. 5, pp. 519-547, 1999).
[10] A. Imburgia, R. Miceli, E. Riva Sanseverino, P. Romano, F. Viola -
Review of space charge measurement systems: acoustic, thermal and
optical methods (IEEE Trans. Dielectr. Electr. Insul., Vol. 23, No. 5,
pp. 3126 – 3142, 2016).
[11] G. Mazzanti, G. Chen, J. Fothergill, N. Hozumi, J. Li, M. Marzinotto,
F. Mauseth, P. Morshuis, A. Tzimas, C. Reed, K. Wu - A protocol for
space charge measurements in full-size HVDC extruded cables (IEEE
Trans. Dielectr. Electr. Insul., Vol. 22, N. 1, pp. 21-34, Feb. 2015).
[12] IEEE Standard 1732, Recommended practice for space charge
measurements in HVDC extruded cables for rated voltages up to 550
kV, IEEE DEIS TC/WG “HVDC Cable Systems”, June 2017.
[13] CIGRÉ TB 636, Diagnostics and accelerated life endurance testing of
polymeric materials for HVDC application, WG D1.23, Nov. 2015.
[14] Brochure CIGRÉ 496 - Recommendations for testing DC extruded
cable systems for power transmission at a rated voltage up to 500
kV - CIGRÉ Working Group B1.32, 2012. [15] IEC Standard 62895
- High Voltage Direct Current (HVDC) power transmission cables
with extruded insulation and their accessories for rated voltages up to
320 kV for land applications - Test methods and requirements - Ed.
1.0, May 2017.
Cigre Science & Engineering • N°13 February 2019
44
[16] T. Maezawa, J. Taima, Y. Hayase, Y. Tanaka, T. Takada, Y. Sekiguchi,
Y. Murata - Space Charge Formation in LDPE/MgO Nano-composite
under High Electric Field at High Temperature (2007 Annual Re-
port Conference on Electrical Insulation and Dielectric Phenomena,
pp.271-274).
[17] T. Mizuno, T. Takahashi, H. Harada, N. Hayashi, Y. Tanaka and
T. Maeno, 2011 - Effect of conductive inorganic filler on space
charge characteristics in XLPE as a HVDC insulation material (8th
International Conference on Insulated Power Cables Jicable’11, paper
No. C.5.2, pp. 1-6, Versailles, France, 19 – 23 Jun. 2011).
[18] M. Marzinotto, G. Mazzanti, U. Vercellotti, H. Jahn - On the way
to compare the polarity reversal withstand capability of HVDC
Mass-Impregnated and extruded cable systems (9th International
Conference on Insulated Power Cables Jicable 15, paper A10.6, pp.
1-4, Versailles, 21-25 June 2015).
Comparative study of the long-term
reliability of HTLS conductor systems
W. KIEWITT**, M. WUNKTE, 50Hertz, Germany
R. BARDL, BAM Berlin, Germany
C. KÜHNEL, IEEH TU Dresden, Germany
D. LOUDON, EFLA, Norway
D. STENGEL, IPU Berlin*, Germany
Summary
In recent years, several high temperature low sag
conductors (HTLS) have been developed and are now
commercially available. Very few utilities have long-term
experience with all the technologies. Moreover, most of
the operational experience is at low temperatures. An
independent testing program at maximum operational
temperatures comparing most of the available technologies
would provide useful guidance to utilities considering
using such conductors.
This paper describes the approach and results of an
international research project assessing the long-term
reliability of a wide range of commercially available high
temperature conductor systems. Each system was exposed
to 4,400 hours of simultaneous mechanical and electrical
stresses. A set of standard mechanical and electrical tests
were carried out to determine any changes. A conventional
ACSR conductor was also tested and provided a validation
that the testing program was realistic.
The majority of the conductor systems performed
satisfactorily. However the ACSS system joint resistance
and the grease on the TACSR performed poorly.
The work was done as part of the BESTPATHS project
(refer to ‘Acknowledgements’).
1. Introduction
The demand for electricity continues to increase, while
operational constraints and (lack of) public acceptance
* Former employee of BAM Berlin until October 2017
** wilhelm.kiewitt@50hertz.com
KEYWORDS
High temperature low sag conductors (HTLS), ageing, reliability, joint resistance, conductor grease, stress-strain,
mechanical and electrical tests
Cigre Science & Engineering • N°13 February 2019
45
make the construction of new overhead lines challenging.
Reconductoring of existing overhead lines with high
temperature low sag (HTLS) conductors provides an
attractive solution to increase the capacity and flexibility
of transmission systems.
Several HTLS conductor technologies are available on
the market, utilizing various conductor and core materials.
A significant amount of research has be done on these
products with different foci, but we still lack a definitive
set of internationally accepted standards for qualitative
testing of these technologies taking into consideration
their special characteristics. Moreover, only a few utilities
worldwide have long-term operational experience with
HTLS conductor technologies, so there is a need for both
documented experiences and independent research to help
utilities to decide which HTLS conductor technology best
suits their needs.
To this aim, an international consortium of research
institutes, utilities and a university carried out a three year
research project on the long-term operational reliability
of commercially available HTLS conductor systems.
The goal was to detect and compare changes of selected
conductor and connection characteristics caused by a
long-term electrical, thermal and mechanical loading
regime. Several mechanical tests were done before and
after long-term loading and electrical performance of the
conductor connections was recorded during the loading.
A conventional ACSR conductor was tested with equal
loading for validation and comparison of the tests and the
loading regime.
 Aluminium Aluminium Aluminium Aluminium Thermal Super Aluminium
Conductor Conductor Conductor Conductor Aluminium Thermal Conductor
Composite Polymer Composite Steel Conductor Aluminium Steel
Core Reinforced Reinforced Supported Steel Conductor Reinforced
ACCC ACPR ACCR ACSS Reinforced Invar ACSRII
TACSRI Reinforced
ZTACIR

Outer material Al99.7 III AT1IV AT3IV AL0V AT1IV AT3IV AL1VI

Core
Core material
material Composite
Composite Composite
Composite Al
Al matrix
matrix ACSHS
HS steel
steel
wires Invar steel Invar steel
ACS wires Galvanized
Core
Core core core
wires wires wires
wires wires wires steel wires

Outer diameter 36.9 mm 36.9 mm 26.4 mm 27.5 mm 27.0 mm 29.3 mm 27.0 mm

Operating temp. 175 °C 150 °C 210 °C 200 °C 150 °C 210 °C 80 °C

Emergency temp. 200 °C 1 180 °C 1 240 °C 2 230 °C 180 °C 1 240 °C 1 -

Rated tensile
strength (RTS) 248 kN 242 kN 113 kN 156 kN 121 kN 239 kN 123 kN

I) High temperature, conventional sagging behaviour V) according to EN 50540

II) Conventional conductor, not HTLS VI) according to EN 50182
III) according to EN 573 1) for up to 400hours
IV) according to EN 62004 2) for up to 1,000 hours

Table I: Specifications of investigated HTLS conductor systems according to manufacturing datasheets

Conductors to a delivery error the wrong conductor samples were

The selection of conductor technologies for the presented
research was based on an analysis of relevant high
temperature conductor technologies available on the
market at that moment with the goal to enable a comparison
of a wide range of alternative conductor technologies.
It is to be noted that only one product of one manufacturer
was investigated for each conductor technology.
Therefore, the presented results may not be applicable
for other products of the respective technology from
other suppliers. Table I presents the selected conductor
technologies together with their key properties as provided
in the manufacturer’s datasheets.
A gap conductor (GZTACSR) was also selected but due
tested and the test results were not usable.
Conductor fittings
Conductor fittings that are subjected to high tensile
stress and electrical loading are equally important to the
long-term reliability of the conductor system. Therefore,
all conductors were tested together with the respective
dead ends and compression joints as proposed by the
manufacturers.
Due to the diversity of the conductor fittings (varying
geometries, materials and mounting technologies) (as
shown in Figure 1), all specimens used for this research
were assembled by trained personnel provided by the
manufacturers.

Figure 1: Comparison of scaled cross-sections of investigated compression joints

Cigre Science & Engineering • N°13 February 2019

46
 Figure 2: a) Picture and b) sketch of test rig for mechanical and electrical long-term loading [3]
2. Long-term loading in the
conductor test rig
The long-term mechanical and electrical loading was done
on a custom built test rig (see Figure 2a [3]). Due to space
limitations, two separate long-term loading sets each with
four conductor systems were carried out sequentially. For
each conductor system two specimens were tested: One
continuous conductor sample of about 20 m length and
a second sample with three compression joints evenly
spaced over the same length (see Figure 2b).
The specimens of each conductor type were connected
in series using jumper lugs. The current was provided by
a separate high-current transformer for each conductor
type and controlled by a variable transformer which
has an automatic temperature control. Furthermore,
the specimens of each conductor type were connected
to a hydraulic cylinder, which was controlled with an
automatic force control. This allowed the application of
additional static and/or cyclic mechanical loads.
Every conductor type was loaded with a current to reach
and maintain the conductor’s rated temperature. The
temperature was measured with thermocouples in the
middle of the continuous specimen of each conductor
type. After 4,000 h of constant temperature, the current
load was raised in order to reach the specified maximum
temperature in emergency conditions for another 400 h.
This 400h period was chosen to account for the lowest
time limitation for emergency operation of the selected
conductors (see Table I).
In parallel to the electrical loading, mechanical loads were
applied to the conductors. A permanent static mechanical
load correlating to the horizontal component of the tensile
force in an ideal single span of an overhead line was
applied. Based on relevant wind velocities, which were
accumulated over an estimated lifetime of about 50 years,
resulting cyclic tensile forces were estimated and applied
periodically to the conductors on top of the static load.
Cigre Science & Engineering • N°13 February 2019
47
The result was a static mechanical load of 15 % rated
tensile strength (RTS) superposed with three different
additional cyclic load levels between 18 and 28 % RTS
depending on the conductor technology (for details see
[2]). Additionally, each conductor was loaded four times
with 50 % RTS, simulating exceptional loads such as ice
loads. Due to the characteristics of the test rig the applied
mechanical loads were independent from the temperature
of the conductor samples.
3. Description of the conductor
tests
In order to assess the reliability of the investigated
conductor technologies together with the fittings a set
of mechanical and electrical tests was carried out and
results were compared to each other. In the following, the
specifications for the executed tests are presented in more
detail.
Conductor creep
The evaluation of creep prediction was carried out only
on the new conductor samples as the rate of creep is so
high during the first weeks of operation. Creep tests were
conducted in accordance with EN 61395. The following
test parameters were used:
Time t = 1,000 hours
Temperature ϑair = 20 °C
Stress H = 20 % RTS
Conductor self-damping
The goal of the self-damping test is the comparison of
dynamic properties of different HTLS technologies before
and after the long-term loading. To detect ageing effects,
it is sufficient to test the conductor at one load (15 %
RTS), instead of three as suggested by the EN 62567. For
this load, ten natural frequencies were estimated for wind
velocities from 1 to 7 m/s. As given by the standard, three
different double amplitudes of 30/2f, 70/2f and 110/2f
were selected for the self-damping test.
 Figure 3: Exemplary test setup for a) single wire, b) ACCC and c) ACPR core tensile strength test
At resonant frequency, a data sequence of one minute was
analysed and the power dissipation was calculated using
the “power method”.
Conductor stress-strain behaviour
The stress strain behaviour has been tested before and after
the ageing test. The aim is the identification of changes in
the modulus of elasticity and breaking strength due to the
long-term loading. Stress-strain tests were conducted in
accordance with EN 50182 and EN 50540.
On one side of the test setup a hydraulic cylinder and a load
cell were mounted in series to tension the conductor with
the required force. Under pre-tension of 5 % a gauge length
of 4 m was installed in the middle of the conductor and a
draw-wire sensor was used to measure displacement. From
the stress-strain data two moduli (below and after knee-
point) were estimated for each conductor.
Conductor grease
As described in EN 50182, the weight of grease in the
conductor was estimated. This test determines losses or
changes of conductor grease due to the long-term loading.
For the test, a piece of conductor (length approx. 1 m) was
cut and weighed. It was disassembled and single wires
were cleaned with a cloth and acetone. To determine the
mass of grease the bundle of cleaned wires was weighed a
second time. The difference between mass of grease for the
new and the aged conductor is the grease loss.
Fittings slippage
The tests were conducted in accordance with EN 61284.
The same test rig as for the conductor stress strain behaviour
was used. By testing the samples, three joints and two dead
end joints were tested simultaneously. Under pre-tension of
15 % RTS, at each interface of a fitting (joint/ dead end
joint) and the conductor, a draw-wire displacement sensor
was installed to measure relative displacement between
conductor and fitting. The draw-wire sensor was fitted on
the conductor with a pair of clamps and connected to a
hook which was glued on the fitting. The maximum load
of the conductors was chosen between 75 % and 85 % RTS
depending on its behaviour in the stress strain test of the
Cigre Science & Engineering • N°13 February 2019
48
continuous conductor to prevent an early breakage and to
protect the measurement equipment.
Single wire tensile strength (conductor core)
Steel wires as well as ACCR core wires were tested in
clamps for round wires or were glued in tubes prior to
testing. The ACCC composite core was tested in the
clamping system as used for the conductor installation and
the ACPR core was glued in epoxy resin (Figure 3).
The specimen was pulled until failure. Tests were conducted
at room temperature. Load and displacement were recorded
and at least three samples were tested per technology.
Single wire tensile strength (aluminium wires)
Wires from the outer layer were tested. To clamp the
specimen in the testing machine, special clamps for round
wires were used. The specimen was pulled with 1 mm/s
until failure at room temperature. Load and displacement
were recorded. At least three samples were tested per
technology.
Single wire wrapping
Single wire wrapping tests were conducted on the
aluminium wires as described in EN 60889 and ISO 7802.
The conductor was cut into pieces of 60 cm, disassembled,
cleaned and the wires were straightened by hand. For the
test, a bolt with the corresponding size was clamped inside
a vice. The round wires were wrapped around a bolt with
the same diameter and formed wires were wrapped around
a bolt of the minimum diameter of the wire. Each wire was
wrapped 8 times around the bolt, unwrapped 6 times and
rewrapped 6 times. Then it was checked for visible cracks.
Joint resistance and conductor DC resistance
The joint resistance Rj of each compression joint and
dead-end is periodically measured taking into account the
given standard for overhead lines EN 61284. The electrical
resistances were measured at room temperature with four-
terminal-sensing using DC current and a microohmmeter.
The potential on the conductor was tapped with a potential
equalizer, the second potential measuring point was located
at the end of the compressed section (Figure 4). The
 Figure 4: Schematic overview of the joint resistance measurement and the equivalent circuit [3]

conductor DC resistance was measured before and after
the ageing tests over a length of three meters. Thereby, the
conductor’s potential is tapped with a potential equalizer,
too. All readings were referred to a temperature of 20 °C
and in case of joint resistances the material resistance of the
25 mm conductor section was subtracted.
Due to the large size of joints for overhead lines and
especially for HTLS, the bulk resistances of the sleeve
and the conductor usually form a large share of the joint
resistance. Thus, the measured resistance is not sensitive
to the contact behaviour of the joints. By modelling the
connections with an equivalent circuit of infinite electric
resistances, it is possible to get a closer look on the contact
behaviour [1]. The specific resistance R’q combines all
contact spots between the single wires of the conductor as
well as between the sleeve and the conductor. Thereby, with
the specific resistance R’q the contact behaviour can be
considered simply without an influence of the connection’s
material properties and geometry.
4. Test results
Overview of the test results
Based on the extensive results of all executed mechanical
and electrical conductor tests, an overview table was
developed in order to summarize the results in a compressed
form (Table II). In the below table a tick mark represents a
passed test (All requirements were met or surpassed). A red
box indicates that the test was failed and the samples did
not fulfil the minimum requirements. A yellow box means
that the requirements were nearly met.

Table II: Overview of executed tests and results before (new) and after long-term loading (aged)

Cigre Science & Engineering • N°13 February 2019

49
 Figure 5: Grease of TACSR: left: new; right: aged, and partially removed

Results for conductor creep tested steel core due to the ageing regime. However, the
composite cores of ACCC and ACCR showed a reduction
While the extrapolated 10-year creep of conductors with
of strength by 7 % (ACPR: not tested). Besides the change
a steel core ranged between 0.027 % and 0.047 %, the
in breaking load, a change in colour and breaking pattern
composite core conductors showed either a similar (ACCR:
was observed for the ACCC core. The tested aluminium
0.035 % and ACPR: 0.032%) or a higher elongation
wires showed a reduction in breaking load of 3 to 5 % due
(ACCC: 0.061 %). The obtained results are distinctive
to the ageing. It is assumed that the cause can be either a
for the tested conductors and different creep rates may be
change of the material properties or fretting of the wires.
found for conductors of different manufacturer. Moreover,
creep rates depend on the conductor properties, like the Results for conductor grease
usage of grease.
Out of the tested technologies, three conductors contained
Results for conductor self-damping grease. Grease needs to meet many requirements (retain
its properties over a high temperature range, no migration,
The tested conductors showed either no effect or an
protection, easy application) [4]. The grease of the TACSR
increase in power dissipation due to the ageing test.
technology did not retained its properties as it became
This can be explained by the mechanical stress applied
harder and drier due to the ageing regime and a grease
to the conductors. Creep and temperature cause plastic
loss of 57 % was detected. It cannot be evaluated whether
deformation, especially of the aluminium wires, while the
this affects the protection of the conductor. Furthermore, a
cyclic loading slightly loosens the stranded wires. This
grease loss of 22 % was observed for the ACPR technology
allows a larger relative movement, higher friction and
and a grease loss of 8 % for the ACSR. The conducted
subsequently leads to higher power dissipation in the aged
experiments show that the selection of grease is a difficult
conductor compared to the new one.
process and a drop point as well as an oil separation test
Results for conductor stress-strain behaviour and should be considered to qualify the grease.
single wire tensile strength
Results for fittings slippage
In the test, no significant changes in the stress-strain
The displacements caused by the slippage test are shown
behaviour of the conductor (breaking strength and
in Table III.
Young’s modulus) were detected. A reduction of breaking
load of the ACSR conductor (see Table II) at the dead ends The displacements of the ACSS and ACCC technology
was detectable but still within the testing requirements. exceeded the standard’s threshold of 2 mm (see Figure
Furthermore, no strength loss was detected for any 6). However, the standard used for this test (EN 61284)

Table III: Results of the fittings slippage test

Slippage
Conductor type Maximum load (RTS)
New Aged
ACCC 80 % Min. 6.8 mm, Max. 11 mm Not tested
ACCR 80 % (new), 75 % (aged) <2 mm <1 mm
ACPR 80 % <1 mm <1 mm
ACSS 80 % Min. 2.4 mm, Max. 4.4 mm <1 mm
TACSR 80 % <1 mm <1 mm
ZTACIR 80 % <1 mm <1 mm
ACSR 75 % <2 mm <1 mm

Cigre Science & Engineering • N°13 February 2019

50
was developed for ACSR which consists of hard drawn
aluminium instead of annealed aluminium. Besides
the aluminium alloy, for the ACCC, the wedge-like
technology which was used to clamp the core is prone to
cause displacement. At the time of writing this document
the standard is reviewed for conductors that use annealed
aluminium.
Figure 6: Slippage at ACCC conductor joint specimen
All tested aged conductors show lower displacements
than the new conductors. This might be due to the applied
forces and subsequently displacement which already
occurred during the long-term loading.
Results for joint resistances
Based on the results given in Figure 7 the connections for
all investigated conductor technologies can be divided
into three different groups.
The connections with the ACSR as well as connections
with the ACPR, TACIR and ACCR can be assigned to the
first group. In general, the specific resistances R´q of those
connections are characterised by very small changes over
time (Figure 7). Based on that stable contact behaviour, a
safe operation of those technologies at normal service life
can be expected.
The connections with the TACSR and ACCC belong to
the second group, which is characterised by a continuous
increase of the specific resistance over time without
reaching a stable level and combined with an increase of
the scattering. This indicates an ongoing deterioration of
the contact behaviour. Compared to the theoretical ageing
behaviour of an electrical connection [1], it is assumed that
the phase of relative stability, determining a connection’s
lifetime, has not been reached during the testing time.
That is one of the main differences to the conductor
technologies from the first group and it indicates a lower
reliability of the connections with TACSR and ACCC over
time. However, over the period examined the observed
changes are not critical yet.
Cigre Science & Engineering • N°13 February 2019
51
The third group is characterised by a huge and critical
increase of the joint resistance, leading to connection
temperatures higher than the conductor’s temperature.
This belongs to the connections with the ACSS.
Compared to the other conductor technologies, these
connections started with the highest joint resistances and
the worst contact behaviour already. Over the period of
examination the averaged specific resistances showed a
deterioration of the contact behaviour by over 16 times
the initial value (Figure 7). However, the large share of
bulk resistance in the measured joint resistance covers
the most of this deterioration. That is why the measured
joint resistances increased by only less than 4 times the
initial values (Figure 7). Based on these results, the long-
term behaviour of these connections clearly needs to be
considered as unstable.
As already concluded in previous investigations, the
residual contact stress at the contact interfaces (single
wires / sleeve and single wires/single wires) significantly
influences the connections quality and reliability over
time [1]. Thereby, the residual contact stress is mainly
influenced by the degree of deformation of the contact
partners during assembly, the mechanical strength of the
aluminium alloys and the connection’s geometry [1].
Due to a lower mechanical strength, the residual contact
stress of connections with annealed aluminium wires are
potentially lower than with hard-drawn aluminium wires
(ACSR, ACPR, ACCR, ZTACIR). In consequence, the
connections with annealed aluminium wires (ACSS,
ACCC) are more susceptible to the applied electrical-
thermal and mechanical stresses (Figure 7). Moreover,
the design of the connections with the ACSS indicates
a comparatively small deformation and compression of
the contact partners (Figure 1), which lower the residual
contact stress even more. This probably caused the
observed significant deterioration of the contact behaviour.
The ACSS and the TACSR show a significant resistance
drop after the period at emergency temperature (Figure 7,
left). During that period a power failure occurred, which
led to a rapid cooling of the specimen. As the hydraulic
pressure of the control unit could not be adjusted without
external energy, this was superposed by an additional
increase of the tensile stress (ACSS: up to ≈ 62 % of the
conductor rated tensile strength). After the restart, all
joints showed lower temperatures than before.
 Figure 7: Joint resistance Rj and specific resistance R´q of investigated compression joints over time
This probably indicates that the exceptional high
mechanical loads caused some relative movements at the
contact interface, leading to enhanced joint resistances
and lower temperatures. Due to their better electrical
contact behaviour and probably higher remaining contact
forces [1], the effect on the joints with the TACSR is not
as remarkable as of the joints with the ACSS.
5. Conclusion
An extensive testing program for high temperature
conductor systems (conductor + fittings) was presented
in this paper. The long-term reliability was assessed by
a combination of a long-term mechanical and thermal
loading with several (diagnostic) tests before and after.
The validity of the testing regime was confirmed by the
satisfactory performance of the reference conductor, a
conventional ACSR.
The investigated ACCR, ACPR and ZTACIR conductor
systems passed all tests before and after long-term loading.
The ACCC and TACSR conductor systems passed all
tests except for the fittings slippage (ACCC) and grease
test (TACSR). The investigated ACSS conductor system
failed with regard to joint resistance and fittings slippage.
In the conducted tests, a reduction of tensile strength was
observed for composite core materials. Future research
is necessary to assess these ageing effects in detail.
Additionally, the influence of corrosion on HTLS at high
temperatures is still subject to research. As observed for
Cigre Science & Engineering • N°13 February 2019
52
some of the test results on joint resistance also a longer
time period for long-term loading might provide a
better basis for extrapolation of test results on long-term
behaviour.
6. Disclaimer / Acknowledgement
The authors acknowledge the key role of “BESTPATHS”
in the funding of this research. “BESTPATHS” stands for
“BEyond State-of-the-art Technologies for rePowering
AC corridors and multi-Terminal HVDC Systems”.
It is co-funded by the European Commission under
the Seventh Framework Programme for Re-search,
Technological Development and Demonstration under the
grant agreement no. 612748.
7. Bibliography
All standards referred to in this publication were used in
the most recent version available before the date of this
publication.
[1] C. Hildmann “Zum elektrischen Kontakt- und Langzeitverhalten
von Pressverbindungen mit konventionellen und Hochtemperatur-
Leiterseilen mit geringem Durchhang” (PhD Thesis TU Dresden,
2016, http://nbn-resolving.de/urn:nbn:de:bsz:14-qucosa-222889).
[2] D. Stengel, et al. “Accelerated electrical and mechanical ageing tests of
high temperature low sag (HTLS) conductors” (Proceedings ICOLIM
conference, Strasbourg, April 2017).
[3] C. Kühnel, et al. “Investigations on the mechanical and electrical
behaviour of HTLS conductors by accelerated ageing tests” (24th
CIRED Conference, Glasgow, June 2017, http://ietdl.org/t/ykwGHb).
[4] Experience with the mechanical performance of non-conventional
conductors, Cigre Working Group B2.48, 2017
Voltage uprating of existing Eskom high
voltage substations when transient
voltage stress and available withstand
strength are coordinated
P.J. Schutte*, Eskom SOC Ltd., South Africa
W.C. van der Merwe**, Eskom SOC Ltd., South Africa
J.M. van Coller***, University of the Witwatersrand, South Africa

Summary When considering the standard approach by IEC and

Uprating the nominal operating voltage from 275 kV
to 400 kV requires an extensive insulation design
evaluation. When the system is operated at an altitude of
1600 m above sea level, even more emphasis is put on
the transient withstand capability of external clearance.
The idea is to define the switching impulse withstand
level (SIWL) as a function of the time to crest that is not
necessarily the critical wave front (CWF). If the specific
gap is tested and the U –curve is well defined, different
critical flashover voltages (CFO) could potentially be
defined as a function of rise time. The Gallet – Leroy
equation defines the clearance as a function of the
critical wave front, but for gaps in the range 2 m – 3 m,
the critical wave front could be in the order of 80 μs. The
expected transients will most likely have rise times much
slower than 80 μs. The withstand characteristic could
be evaluated to derive more relevant critical flashover
values that relates to longer rise times in the order of
200 μs and longer. To match the withstand characteristic
with the expected stress, the impact of the network
configuration on the time to crest were evaluated.
When considering the expected over-voltage wave shapes,
this paper also demonstrates that for short lines, the natural
frequency transient oscillation (after energisation) is well
interrelated with the time to crest of the initial transient
measured at the mid-point of the line. For longer lines
standing waves dominates the impulse response and the
natural frequency can’t be related to the first time to crest.
For shorter lines, the expected rise times of the switching
surge can be estimated. The rise times as a function of
transmission line length and short circuit power can
contribute to define the stress in more detail.
IEEE, the lowest CFO will be selected for the risk
calculation to ensure conservativeness. The case study
proved that for the SIWL of to 770 kV, based on the
critical wave front of 84 μs, the risk was calculated to
be 0.8 flashes per year related to 100 events per year.
This value would be considered to be high and a value
of 0.1 flashes per year (100 events per year) should be
more acceptable. The risk value is significantly reduced
to the desired value when the SIWL of 830 kV is chosen.
When the SIWL of 830 kV (value related to withstand
characteristic of 200 μs and more) is used with the idea
to integrate the withstand characteristic, the calculation
yields an acceptable risk of 0.11 flashes per year and is
relevant to the expected stress as defined by the network
configuration and transmission line details.
1. Introduction
With the increasing expansion requirements for power
systems in built up areas, voltage uprating is often
considered as a means to utilise existing transmission
infrastructure and increase power transfer capabilities.
In this case the transmission system currently operating
at 275 kV is studied to be uprated to 400 kV at an altitude
of 1600 m above sea level. The insulation design of
the current substation and transmission line clearances
becomes crucial as we know that larger clearances are
sensitive for switching surges [1].
The standard approach of switching surge insulation
coordination includes the electromagnetic transient
statistical simulation to determine the over-voltage
distributions [2]. A standard switching impulse withstand
level is selected and the risk of failure is calculated.

* schuttpj@eskom.co.za
** vdmerwwc@eskom.co.za
*** john.vancoller@wits.ac.za

KEYWORDS
Insulation co-ordination, switching surge, withstand level, voltage uprating, clearance, over-voltage, transient, critical
flashover voltage, critical wave front.

Cigre Science & Engineering • N°13 February 2019

53
Usually statistical safety factors are included to ensure
that the risk of failure is minimised. The selected
switching impulse withstand level for external insulation
is well defined by the Gallet - Leroy equation [3] as a
function of clearance and gap factors. It is clear that
the equation is based on the critical wave front for each
gap which yields the lowest flashover voltage on the
withstand characteristic. For gaps between 2 m and 3 m,
the critical wave front could be significantly different to
that of “large gaps”. The critical wave front is expected
to be in the order of 200 μs to 250 μs, depending on
the gap factor and clearance. But when the critical wave
front of a 2.3 m gap is in the range of 80 μs, the wave
front is considered to be very fast for a switching surge.
It is also considered not fast enough for a lightning surge.
Therefore, to evaluate the gap as part of the “large gaps”
group will not yield representative results. If the critical
flashover voltage at the critical wave front is correlated
to expected stresses, it is justified to use this value. But if
the network configuration only allows for expected rise
times of 200 μs and longer, the representative critical
flashover voltage can be used to represent the expected
stresses. This value will then yield a more realistic risk
of failure. Environmental factors such as humidity and
altitude has a significant impact on the withstand level
of external insulation and should be included in the
calculation.
2. Withstand level as a function
of the switching impulse
withstand characteristic with
varying time to crest value
it is well-known that apparatus-based electrical clearances
have been verified by extensive laboratory testing and
have existed for several years. Due to the complexity
Table 1 : Standard clearances [4]
Nominal
Nominal Withstand Levels (kV)
Maximum
Voltage
Voltage
(RMS phase-
(RMS phase- LIWL SIWL
to-phase kV)
to-phase kV)
275 300 1050 850
400 420 1425 1050
Cigre Science & Engineering • N°13 February 2019
54
of substations as whole, typical combinations of
hardware assemblies, clamps, bus conductors, conductor
stringers, droppers and insulators, have not been tested
as extensively. For this application, the concept of
an uprated 275 kV transmission line is evaluated. An
existing 2.32 m phase-to-earth airgap is usually adequate
for a system voltage of 275 kV, but careful consideration
should be taken to determine how stress and strength
are co-ordinated in terms of the transient phenomena in
an uprated system. The standard withstand levels and
clearances used in the Eskom grid is given in Table 1.
To evaluate the electrical clearance for the typical
2.35 m gap utilised in the 275 kV system in South
Africa, impulse tests resulted in the determination of
the withstand capability of this gap. The tests were
performed for an emergency uprating project and the
same results are used in this paper. Four impulse values
were chosen to produce the typical U-curve, namely
4/56, 84/1928, 130/2040 and 260/2208 μs. The up and
down method of testing was used to determine the U50
value for each impulse type with 30 shots per impulse.
When considering the standard approach to calculate
the risk of failure, a limitation exist (when considering
uprating of existing infrastructure) of co-ordinating
the stress and strength for smaller gaps. Taking the
calculated switching impulse withstand level and
selecting (based on previous experimental data) the gap
factor at an altitude of 1600 m will inherently include a
significant error. Furthermore, the calculated withstand
level is only a function of the strike distance and the
gap factor referenced to the critical wave front. A more
valuable approach for smaller gaps is to subject the
specific airgap to an impulse test at altitude to obtain
the specific withstand characteristic (U – curve) and
then determine the total sensitivity towards the time to
Minimum Electrical Minimum Working
Clearance (mm) Clearance (m)
Phase-to- Phase-to-
Vertical Horizontal
earth phase
2350 2950 4.8 3.4
3200 4000 5.7 4.3
 Figure 1: Risk calculation based on a withstand level as a function of the critical time to crest
crest and correlating that to the expected stresses. The
U – curve could potentially be well defined and provide
information regarding its sensitivity when subjected to
faster front switching surges as opposed to slower front
switching surges. The sensitivity can be translated into
risk areas, where different critical flashover voltages are
defined.
In Figure 1 (U –curve from actual test data for 2.32 m
gap), the SIWL is specified as a function of the time to
crest. The time to crest in the range of 40 μs - 200 μs,
yields a corresponding value for the critical flashover
voltage of 822 kV (red area and the lowest CFO obtained
during 80 μs time to crest). For times to crest exceeding
the 200 μs, the lowest CFO of 979 kV is selected (blue
area). The two different regions relate to two different
SIWL values and will yield different results when the
risk calculation is performed.
The gap critical flashover voltage is defined as a function
of the critical time to crest and represented by the U - curve.
With this varying parameter (time to crest) included in
the risk calculation the withstand level of the gap could
be adjusted as per the expected time to crest of switching
surges. It is found that the network configuration and
the transmission line parameters directly determines the
expected rise times of the switching surges.
This paper demonstrates that for short lines, the natural
frequency transient oscillation (after energisation) is well
interrelated with the time to crest of the initial transient
measured at the mid-point of the line. For longer lines
standing waves dominates the impulse response and the
natural frequency can’t be related to the first time to crest.
For shorter lines, the expected rise times of the switching
surge can be estimated. The rise times as a function of
transmission line length and short circuit power can be
related to the withstand characteristic. The value for the
switching impulse withstand level is then selected based
on this calculation and the risk calculation is expected
to yield more relevant results when compared to the
standard approach.
Regarding transmission line capacity upgrading, a
vast amount of literature is available. The methods
Cigre Science & Engineering • N°13 February 2019
55
and opportunities available to upgrade the capacity of
transmission lines are well described in the literature [5]
[6]. The critical flashover voltage is referenced to the
CWF which relates to the lowest flashover voltage on
the U – curve. Instances where the capacity upgrading
in transmission lines has been evaluated proved that
thorough insulation coordination studies are required to
establish the feasibility of voltage uprating [7]. This paper
deals with a gap of 2.32 m phase-to-earth at an altitude of
1600 m above sea level, and therefore demands a novel
approach to evaluate the insulation design. Usually the
substation is not the primary concern when evaluating the
capacity upgrading. Substation upgrading or uprating is
usually triggered by the transmission line infrastructure
constraints. Transmission line uprating or upgrading
methods are well understood and documented [8].
Substation voltage uprating is not so well documented
in the literature. The safety aspects [9] and performance
[10] of uprated substations has been documented but the
literature lacks the design experience. It is expected that
for switching surges, the over-voltage protection at the
substation is adequate to ensure a specific safety margin.
With the standard of implementing line entrance and
transformer surge arresters, switching over-voltages are
expected not to be a problem.
The withstand characteristic in Figure 2 is well defined
and it is clear that the specific air gap is vulnerable to
breakdown when subjected to a time to crest between
50 μs and 200 μs. This area is defined as a high risk
area with medium level of occurrence. The CFO is at its
lowest values in this area and remains within the range
of 820 kV (CFO at the critical wave front) + 50 kV. It
is expected that the transient rise times is a function of
the network configuration and rise times of switching
surges shorter than 50 us is highly unlikely. Rise times
exceeding 200 us will be common and therefore refer
to the high occurrence level area. The observation is
then made that the U – curve can be subdivided in risk
areas. For example, the high risk area represents the
lowest values of the critical flashover voltage but the
occurrence level of surges with relevant rise times is
low. The argument develops now and the risk assessment
 Figure 2: Withstand characteristic for phase-to-earth 2.32 m gap at 1600 m above sea level.
could be potentially approached with the integration of
the risk areas as an additional parameter.
It is then required to establish the possibility of the
power system to generate switching transients with
the critical rise times associated with the different risk
areas. The statistical SIWL can be described by the CFO
as a function of the varying time to crest [11], where
tc represents the four (4 µs, 84 µs, 1130 µs, 260 µs)
different times to crest. Sigma, σƒ is the coefficient of
variation. Sigma is selected to be 5% [11]. The CWF is
84 us and corresponds to the lowest flash over voltage
CFOCWF.
(2.1)
It is clear that the U – curve is required to well define
as a U shape. Flat withstand characteristics represents
unrealistic large values for the gap factor. Gap factors
closer to 1.0 will most likely produce a very well defined
U – curve with clear areas of where the gap will be
weaker and stronger in terms of critical time to crest.
To determine the specific gap factor of the of 2.32 m
gap, altitude correction is considered and the gap factor
is adjusted to obtain the CFO value of 822 kV.
Atmospheric corrections factors are calculated based
on Hilleman’s method [11]. The absolute humidity was
recorded to be 7.9 g/m3 at an altitude of 1600 m. The
normalised CFO for sea level is then calculated to be 1005
kV. The correction factor is then calculated to be 0.865.
This yields a corresponding CFO of 822 kV at 1600 m
above sea level. The gap factor is then adjusted to obtain
the SIWL. The gap factor for phase-to-earth is calculated
to be 1.23. It is estimated that the phase-to-phase gap
factor should be in the range of similar values indicated
for conductor-to-conductor arrangements and selected to
Cigre Science & Engineering • N°13 February 2019
56
be 1.56 to ensure a SIWLphase-to-phase/SIWLphase-to-earth ratio of
1.5. This results in a phase-to-phase SIWL of 1125 kV. The
gap factor of 1.23 is expected to yield a well pronounced
U – curve and explains the well-defined profile of Figure 2.
3. Finding a correlation between
the switching surge time to crest
and the transient response
The primary aim of this paper is to demonstrate that
the risk of failure can be further defined by including
a varying time to crest obtained from a withstand
characteristic (U-curve) of an air gap. This calculation
considers the switching transient wave shape as a function
of the network configuration and the transmission line
geometry.
The concept of relating the switching surge time-
to-crest to the source impedance, transmission line
surge impedance and/or equivalent positive sequence
capacitance is further explored.
When energising a capacitor through inductive source
impedance, a DC step voltage input (representing the
value at a specific time such as a peak of a sinusoidal
power frequency voltage) results in the overshoot and
natural frequency oscillation. In a lumped model the
natural frequency as a function of the source inductance
and the capacitance is present from the instant of the step
response initiation. In the lumped model, the propagation
of transients is not included. The response includes
therefore purely the oscillation as described by the
natural frequency. The rate of rise of the initial overshoot
(switching surge) is derived from the natural frequency.
The time to crest is estimated to be ¼ of the sinusoidal
period of the waveform. The oscillations are present
within the transient duration and the after the oscillation
is fully damped only the original source voltage remains.
 Figure 3: Long line response (200 km)
It is therefore expected that a strong relationship exists
between the time to crest and the natural frequency of
the source inductance and capacitance.
The distributed transmission line consists of defined
surge impedance, DC resistance (per meter length) and
a propagation velocity. Switching in the distributed
transmission line is significantly different to that of
a lumped model. At the instant of energisation, the
transient is initiated. This charge up time is a function
of the transmission line parameters as defined by the
surge impedance and the source surge impedance. The
rate of rise and time to crest is related to the charging
up time. Furthermore, depending of the propagation
velocity and the transmission line length, the travelling
wave phenomena will have a significant impact on the
continuity of the initial charge up transient, and the
added reflection.
Short lines with a large surge impedance will generate
short charging times for the initial transient and will
receive the reflection at the sending end almost at the
end of charge up or even during charge up time. This
allows for close to continuum rate of rise of the combined
charging transient with the superimposed reflection. This
continuum slope of the transient may appear to coincide
with the slope of the natural frequency wave shape and
then excite the natural frequency. It is therefore expected
that for short lines with large surge impedances, the rate
of rise and time to crest of the switching surge can be
related to the source impedance and transmission line
capacitance which in turn relates to the natural frequency.
Figure 4 shows that the natural frequency oscillation and
the transient response are very similar in the first time
to crest.
It should be noted that for long transmission lines (with the
same geometry as the short line), exactly the same charge
up transient exists when compared to short lines. This is
due to the charge up transient being a function of the surge
impedance and source impedance and not the line length.
However, the length of the line and the propagation velocity
significantly influences the level of discontinuity of the
total surge (charge up and reflection component).
Cigre Science & Engineering • N°13 February 2019
57
Figure 4 : Short line response (20 km)
Figure 3 represents a long transmission line and the
standing wave dominates the transient response, the
natural frequency does not exist during this time. The
observation is made that the reflection and refraction
of the travelling wave dominates and it is therefore
impossible to relate the switching surge rate of rise to
the natural frequency. It is then clear that the concept
of relating the switching surge to the natural frequency
is only valid for strong sources (small inductance), and
short lines with large surge impedances. Fast travelling
waves and short durations for the reflection at the
receiving end to reach the sending end within the charge
up time is the key. This configuration will lead to an
early excitation of the natural frequency which in turn
can be related to surge steepness. The surge steepness
can be assessed in terms of the withstand characteristic
of the airgap.
Figure 5 presents the concept of calculating the rate of
rise of the swithcing transient (at the mid-point of the
line) as a function of a pre-determined surge impedance,
line length and source impedance. The red and blue
regions relates to the specific withstand characteristic
of Figure 2. As per Figure 5, the blue arrows indicates
the expected transient rise time for a 30 km transmssion
line, energised from a source with short circuit power of
20 kA = 325 μs (which is in the blue region). This value
is confirmed in ATP to be 340 μs.
4. Including the time to crest
within the risk calculation
The transmission line is modelled with the distributed
parameter frequency independent transmission line
model. The geometries of the 275 kV structures were
included to calculate the line parameters. The conductor
bundle includes 4 x Tern stranded aluminium conductors
which are common in the Eskom transmission system.
The surge impedance is calculated to be 237 Ω. For
400 kV ground clearances, the minimum height of the
conductors should be no less than 8.1 m. The sag of
the conductors is correlated with this minimum value.
The transmission line positive sequence capacitance is
 Figure 5: Transient switching surge front time as a function of short circuit power

calculated with ATP line constants to be 12.2 nF/km. To

demonstrate the difference in the standard approach a
case study is evaluated. The transmission line length is
30 km with the single phase source short circuit power
20 kA which yields an expected rise time of 325 μs as per
Figure 5. The source is represented as purely inductive
with an additional resistive load to include damping. The
local busbars are represented with a capacitance of 10 nF
x 400m = 0.004 μF. A statistical switching study of 100
events produces a cumulative over-voltage distribution
at the mid-point as per Figure 7. The U2% value is derived
from this distribution and is estimated to be 1.92 p.u
(1 p.u is 343 kVphase-to-earth peak). Figure 6: 275 kV Transmission line geometry

Figure 7: 400 kV Interconnected transmission lines and substations

Cigre Science & Engineering • N°13 February 2019

58
 Figure 8 Risk calculation based on a SIWL of 770 kV and 830 kV
4.1 Substation switching surge over-voltage
The surge arrester maximum residual voltage for 3kA
switching surge is 657 kApeak. With the inclusion of the
sending and receiving end surge arresters the maximum
over-voltage will not exceed this value significantly.
Based on a SIWL of 770 kV and the protective level of
657 kV, the calculated safety margin at the substation is
14.6 %. This takes into account the calculated gap factor
of 1.23. It is expected that the voltage at the substation
will be controlled with surge arresters, but when
considering the transmission line midpoint, the safety
margin of 14.6% will be reduced due to an increased
over-voltage at the mid-point of the line.
4.2 Transmission line switching surge over-voltage
In an effort to establish the midpoint over-voltage
distribution, the U2% values were simulated with
a statistical ATP analysis. Trapped charge was not
considered for the calculation. To simplify the derivation
of an increased withstand level as a function of switching
surge time to crest, the following assumption is made:
• Transient time to crest between 50 μs and 250 μs =
SIWL 770 kV
• Transient time to crest 250 μs and greater = SIWL
830 kV
In Figure 8 the risk evaluation criteria is described. The
strength or withstand distribution is represented by the
SIWL probability distributions P1(U) = SIWL(830 kV)
and P2(U) = SIWL(770 kV). The risk of failures R1(U)
and R2(U) are calculated with the following integral [2]:
Cigre Science & Engineering • N°13 February 2019
59
(4.1)
The calculation results in R1 (830 kV) = 1.1 x 10-3 and R2
(770 kV) = 7.8 x 10-3. The difference in risk is significant
and will have a notable impact on the feasibility
evaluation of whether or not the insulation will be
adequate. Literature [12] indicates that a value of 0.1
flashes per year (for 100 events per year) is acceptable.
This indicates that the SIWL of 770 kV will yield an
acceptable risk, but increasing the value to 830 kV as per
stress and strength co-ordination, the risk reduces to an
acceptable value of 0.1 flashes per year for 100 events. It
is therefore clear that when the withstand characteristic
is defined as a function of the time to crest obtained from
the U – curve correlated with the expected stresses may
result in a reduction in the estimated risk. Using a higher
value for the SIWL is justified when the expected stress
is correlated with the available withstand characteristic.
5. Conclusion
Uprating the nominal operating voltage from 275 kV
to 400 kV requires an extensive insulation design
evaluation. When the system is operated at an altitude of
1600 m above sea level, even more emphasis is put on
the transient withstand capability of external clearance.
The idea is to define the SIWL as a function of a time to
crest that is not necessarily the critical wave front. If the
specific gap is tested and the U –curve is well defined,
different critical flashover voltages could potentially be
defined as a function of rise time. The Gallet – Leroy
equation defines the clearance as a function of the
critical wave front, but for gaps in the range 2 m – 3 m,
the critical wave front could be in the order of 80 μs.
The expected transients will most likely have transient
rise times much slower than 80 μs. The withstand
characteristic could be evaluated to derive more relevant
critical flashover values that relates to longer rise times
in the order of 200 μs and longer. To match the withstand
characteristic with the expected stress, the impact of
the network configuration on the time to crest were
evaluated.
When considering the expected over-voltage wave
shapes, this paper also demonstrates that for short
lines, the natural frequency transient oscillation (after
energisation) is well interrelated with the time to crest of
the initial transient measured at the mid-point of the line.
For longer lines standing waves dominates the impulse
response and the natural frequency can’t be related to
the first time to crest. For shorter lines, the expected rise
times of the switching surge can be estimated. The rise
times as a function of transmission line length and short
circuit power can contribute to define the stress in more
detail.
When considering the standard approach by IEC and
IEEE, the lowest CFO will be selected for the risk
calculation to ensure conservativeness. The case study
proved that for the SIWL of to 770 kV, based on the
critical wave front of 84 μs, the risk was calculated to
be 0.8 flashes per year related to 100 events per year.
This value would be considered to be high and a value
of 0.1 flashes per year (100 events per year) should be
more acceptable. The risk value is significantly reduced
to the desired value when the SIWL of 830 kV is chosen.
When the SIWL of 830 kV (value related to withstand
characteristic of 200 μs and more) is used with the idea
to integrate the withstand characteristic, the calculation
yields an acceptable risk of 0.11 flashes per year and is
relevant to the expected stress as defined by the network
configuration and transmission line details.
Cigre Science & Engineering • N°13 February 2019
60
6. Acknowledgement
The author acknowledges Eskom Power Delivery
Substation Engineering for the opportunity to participate
in the Eskom Power Plant Engineering Institute (EPPEI)
programme. The completion of this work could not have
been possible without the expertise of Mr Chris van der
Merwe who is a co-author of this paper. The University
of the Witwatersrand and co-author Dr. J.M. van Coller
is acknowledged.
7. Bibliography
[1] CIGRE Working group 07, Study Committee 33, “Guidelines for
Evaluation of the Dielectric Strength of External Insulation,”CIGRE
Technical Brochure 72, 1992. .
[2] IEC 60071-2 (1996), Insulation coordination application guide. .
[3] Gallet, G., et al., “Switching impulse results obtained on the outdoor
testing area at Renardieres,” IEEE Transactions on Power Apparatus
and Systems, vol. PAS-95, pp. 580–585, Mar./Apr. 1976.
[4] A.J.S Groenewald, ‘Eskom Substation layout design guide, Insulation
coordination and electrical clearances’. 2011.
[5] I. Albizu, A. J. Mazon, and I. Zamora, ‘Methods for increasing the
rating of overhead lines’, in Power Tech, 2005 IEEE Russia, 2005, pp.
1–6.
[6] K. Kopsidas and S. M. Rowland, ‘Evaluating opportunities for
increasing power capacity of existing overhead line systems’, IET
Gener. Transm. Distrib., vol. 5, no. 1, p. 1, 2011.
[7] K. Kopsidas, S. M. Rowland, M. N. R. Baharom, and I. Cotton,
‘Power transfer capacity improvements of existing overhead line
systems’, in Electrical Insulation (ISEI), Conference Record of the
2010 IEEE International Symposium on, 2010, pp. 1–5.
[8] J. R. Daconti and D. C. Lawry, ‘Increasing power transfer capability
of existing transmission lines’, in Transmission and Distribution
Conference and Exposition, 2003 IEEE PES, 2003, vol. 3, pp. 1004–
1009.
[9] S. V. UPRATING, ‘Safety aspects in substation voltage uprating’,
IEEE Trans. Power Delivery., vol. 7, no. 3, 1992.
[10] W. J. Lannes, H. Schneider, J. A. Jones, and M. Spalding, ‘Performance
of uprated substations in contaminated environments’, IEEE Trans.
Power Deliv., vol. 13, no. 4, pp. 1143–1156, 1998.
[11] A. R. Hileman, Insulation coordination for power systems. CRC
Press, 1999.
[12] L. Thione, ‘Insulation coordination in electrical power systems thery
and application’. Aug-2001.
Research and development of
Ultra-High-Voltage VSC for
the multi-terminal hybrid ±800kV HVDC
project in China Southern Power Grid
H. RAO, Y. ZHOU*, S. XU, Z. ZHU
Electric Power Research Institute (State Key Laboratory of HVDC), CSG China

Summary been constructed and put into operation in CSG’s network,

among which eight LCC inverters are located within a
Line commutated converter based high voltage direct 200 km×200 km area in Guangdong Province as shown
current transmission (LCC-HVDC) has been used to in Fig. 1. These eight LCC-HVDCs supply up to 27.2 GW
transfer electricity from bulk power generation to the power in total, which takes up about 20% of the total load
load centre through long distance overhead lines. With in the area. The multi-commutation failure has been the
the increasing use of LCC-HVDC, multi-infeed HVDC very risk needs to be prevented for quite a long time. In
system is applied by companies such as China Southern 2017, the 11th project, the Dianxibei ±800kV/5000MW
Power Grid (CSG). But the multi- commutation failure is LCC-HVDC project, has been put into operation. Its
a great risk needs to be prevented. Many methods have inverter station is also located in Guangdong Province.
been studied to reduce the risk, such as the use of static Consequently, the risk of CSG’s network caused by multi-
synchronous compensator (STATCOM) and synchronous commutation failure has been raised to a new level.
condenser (SC). The voltage sourced converter based
HVDC (VSC-HVDC) provides a new solution for multi-
infeed HVDC systems, since it has no commutation failure
and has flexible reactive power output performance. The
VSC-HVDC can help the power recovery of LCC-HVDCs
by supporting reactive power to power system when faults
occur, which is beneficial to enhance the stability of multi-
infeed HVDC systems. However, the dc voltage, power
rating and topology of current VSC-HVDCs cannot meet
the requirement of long distance bulk power transmission
through overhead lines. Therefore, the VSC-HVDC needs
to be further studied. This paper describes the research
and development of ±800kV multi-terminal LCC-VSC-
HVDC in CSG, mainly focusing on the system circuit Fig. 1. The long distance bulk power HVDC projects in CSG

scheme and VSC topology.
1. Introduction
The network of China Southern Power Grid (CSG) is
characterized by paralleling AC and DC transmission. Up
to now, ten line commutated converter based high voltage
direct current transmission (LCC-HVDC) systems have
In the next several years, the CSG will implement the
12th HVDC project to deliver the western hydropower
of Wudongde generation in Yunnan Province to the load
center in Guangdong Province and Guangxi Province.
The project is named as Wu Dongde(WDD) HVDC,
whose concept is shown in Fig. 2. The WDD HVDC is
planned as a multi-terminal one, and rated at ±800 kV.

* zhouyb@csg.cn

KEYWORDS
Line Commutated Converter, Voltage Sourced Converter, High Voltage Direct Current Transmission, Ultra-high
Voltage, Overhead Lines.

Cigre Science & Engineering • N°13 February 2019

61
 Fig. 2. Possible configurations of WDD HVDC
The rectifier side in Yunnan Province will send 8000MW
in total. The two inverters located in Guangdong Province
and Guangxi Province will receive 5000MW and
3000MW respectively. Meanwhile, the Guangxi side is
required to be able to send power from Guangxi Province
to Guangdong Province. Thus, the total power provided
to Guangdong Province by HVDC transmission will
increase to 37.2 GW totally, which will account for 30%
of the total load in the area.
Based on the existing experience, LCC-HVDC technology
seems to be the only solution for the WDD HVDC.
However, if LCC inverter were once again applied to the
Guangdong side, the multi-infeed effective short circuit
ratio (MIESCR) of WDD HVDC project will be as low as
1.87, which would be too low for LCC-HVDC to operate
safely. In addition, it will cause the MIESCR of the other
nine existing LCC-HVDCs to decrease. Therefore, for
the stability of the system, the solution of installing static
synchronous compensator (STATCOM), is taken into
consideration.
Fortunately, with the development of power
semiconductors and voltage sourced converters (VSC),
especially with the progresses achieved by 3000A
insulated gate bipolar transistor (IGBT), VSC becomes
a possible and feasible solution for the inverter side of
WDD HVDC. The use of VSC at Guangdong Province
will bring at least the following advantages:
• It can secure constant power transmission with no
commutation failure.
• It is able to supply reactive power to the power system
when ac faults occur.
• It helps to reduce the ac fault area that induces multi-
commutation failure.
• It helps improving the voltage recovery process after
ac fault clearance.
• It can help the LCC-HVDC nearby to restore its
power.
Therefore, in order to reduce the multi-commutation
failure, the application of VSC at the Guangdong Province
and Guangxi side is studied while the Yunnan side still
uses LCC. This solution is named as ultra-high voltage
Cigre Science & Engineering • N°13 February 2019
62
hybrid multi-terminal HVDC (UHVH-MTDC). The
following of this paper will discuss this solution.
2. Multi-terminal configuration
With the development of LCC and VSC technology, feasible
multi-terminal configurations for WDD HVDC are shown
as Fig. 2, where four possible configurations are illustrated.
Both Guangxi side and Guangdong side use VSC or LCC,
or one of the Guangxi side and Guangdong side uses LCC
while the other one uses VSC. For scheme II, III and IV, the
LCC based inverter will experience the commutation failure
when ac faults occur. The commutation failure will cause
the dc voltage to drop sharply. Then, it will lead to power
impact to the other inverter, no matter which technology it
uses. Therefore, to some extent, the commutation failure
caused by LCC based inverter enlarges the affected area of
ac faults. Moreover, if the communication is unavailable in
this case, the fault identification of the ac and dc faults for
the other inverter will be a problem.
Since the rectifier has the biggest capacity and the LCC
based inverter has smaller capacity, the coordination
between rectifier and inverter is a great challenge,
especially in occurrence of ac faults and dc faults. Proper
power recovery strategies and coordinated parameters of
voltage dependent current order limitation (VDCOL) must
be designed.
By contrast, if both of two inverters use VSCs, better
performances are available. Firstly, the VSC has no
commutation failure, so the dc voltage will not collapse
when ac faults occur. The ac faults will only influence the
local VSC’s output, leaving the other VSC unaffected.
Secondly, due to the independent control of active and
reactive power, the VSC can release reactive power to the
ac system to assist the recovery process. To tackle with the
fault of overhead lines (OHL), topology of VSC should be
improved, which is discussed in section 4.
Overall, Scheme I, that both of the two inverters use VSCs,
has the best performance, and is the best to ensure the
stability of ac system.
 Fig. 3 Configuration of single converter per pole Fig. 4 Configuration of double converter per pole

3. Schematic of UHVH-MTDC
In terms of system configuration, like the other LCC
stations rated above ±800kV/5000MW, the rectifier side
uses typical configuration, which is double 12-pulse
series-connected wiring, where each one is rated at 400kV
as shown in Fig. 3 and Fig. 4.
For the VSCs, two different schemes can be used. One
is single converter per pole, as shown in Fig. 3, the other
one is the double converter per pole shown in Fig. 4. Both
of the two schemes are compared as shown in table 1. As
can been seen, the VSC with single converter needs fewer
arm reactors, busings, measuring equipment and C&Ps.
Although it needs the same number of transformers as
the VSC with double series-connected converter, all the
transformers are rated at 800kV. Overall, both of the
two schemes nearly cost the same in terms of device
investment, and the footprint is close as well.
As shown in table. 2, the main differences between the
two schemes in terms of operation mode are the converter
switch-in or out. It should be noted that for VSC with
single converter, the VSC itself does not has the demand
of converter switch-in or out. When one of the two
12-pulse converters of the rectifier needs switch-out, the
VSC should work with 50% reduced dc voltage to match
the 400 kV operation voltage. In this case, special VSC
topology is needed, which is discussed in section 4. For
the VSC with double series-connected converter, the VSC
topology only influence the detailed process of converter
switch-in and out.
In terms of control performance, the control mode and
steady-state performance of the two schemes are nearly the
same. The VSC with double series-connected converter
must add the dc voltage balance control between the
high and low converter, when the VSC works at power
control. And its transient performance is influenced by
the parameter difference between high and low converter,
such as the redundant sub-modules quantity. For the
VSC with single converter, in order to get high energy
availability, the VSC should be able to work when one

VSC with double series-connected

Items converter
VSC with single converter
High side Lower side
Capacity/MVA same same same
Transformer Operation two parallel no parallel no parallel
Quantity 12 (rated at 800kV) 6 (rated at 800kV) 6 (rated at 400kV)
Sub-module’s
total half half
quantity
Valve
Terminal to ground 1800 kV 1800 kV 1300 kV
insulation 1600 kV 1600 kV 1050 kV

Terminal to terminal 1300 kV 1300 kV 1300 kV

insulation 1050 kV 1050 kV 1050 kV
Arm reactor Quantity 12 12 12
Quantity (800kV) 6 6 0
Bushing Quantity (400kV) 0 6 6
Quantity (120kV) 3 3 3
Measuring signals About 80 About 110 totally
Control and protection 2 sets 4 sets
Tab.1 Comparison of VSC schemes in main device

Cigre Science & Engineering • N°13 February 2019

63
 VSC with single VSC with double series-
Items converter connected converter

Bipolar YES YES

Monopole (metallic return) YES YES
Monopole (ground return) YES YES
Operation
mode Reduced voltage YES YES
Converter switch-in/out NO YES
Control mode Same Same
Steady performance Same Same

Influenced by parameter
difference between high and low
Transient performance ---
converter

Control
performance Control parameters
need adapt to unavailability of Need voltage balance control
Additional control one transformer between high and low converter

Tab.2 Comparison of VSC schemes in control

of the parallel transformer is unavailable. In this case, the
control parameters should be adaptive.
Overall, the VSC with double series-connected converter
matches the LCC rectifier in terms of wiring, and is more
flexible. Therefore, it is proposed to be applied finally.
In terms of DC yard, for the sake of locating the DC line
fault, a dc bus is designed in Guangxi station. Meanwhile,
in order to fulfill the smooth connection and disconnection
of the two VSC inverters, high speed switches (HSS) are
equiped at the dc poles. In order to fullfill the transition
from ground return (GR) to metallic return (MR) or from
MR to GR, metallic return transfer breaker (MRTB) and
ground return transfer switch (GTRS) are equipped in
the two inverter stations while the rectifier side is not
planned. To suppress the rising rate of fault current of the
VSC valves, dc reactors are equipped. Meanwhile, the
arm reactors are located near the VSC’s dc side.
4. Topology of VSC using
overhead lines
The VSC topology is another key issue of the UHVH-
MTDC. Nowadays, nearly all the VSC- HVDC uses the
modular multilevel converter (MMC) topology because
of its modular design and high output performance. So the
following discussion is based on MMC. On one hand, the
MMC must be capable of providing fast dc fault clearance
by itself to improve the project’s reliability. On the other
hand, it should be able to operate at reduced dc voltage,
such as 80%, 70% or even 50%, in order to meet some
special operational requirements. For the sake of clearing
the dc fault, various MMC typologies have been studied
[1-2] , such as full- bridge sub-module (FBSM), clamped
double-SM (CDSM), clamped single-SM (CSSM), etc.,
as shown in Fig.5.
Among the existing MMC typologies with dc fault

Fig. 5. VSC typologies with dc fault clearance ability

Cigre Science & Engineering • N°13 February 2019

64
clearance ability, most of them clear the dc fault by
blocking the converters except for the FBSM. To some
extent, blocking the MMC will increase the time spent
on clearing the dc fault. This is unexpected. The FBSM-
MMC can operate from -1.0 p.u. to 1.0 p.u. dc voltage.
This characteristic makes it has nearly the same output
performance at the dc side as the LCC inverter, and is
very helpful in improving the operation flexibility and
energy availability. Moreover, the FBSM-MMC can be
able to clear the dc fault by continuously operating at zero
voltage or even negative voltage.
However, in some applications, the power is delivered
unidirectionally, and power flow reversal is not necessary.
In this case, the FBSM-MMC can be further optimized.
That is to say, some FBSMs can be replaced by half-
bridge SMs [3]. This can be named as the hybrid MMC
topology. The hybrid MMC’s available dc voltage depends
on the ratio of FBSM directly, as shown in table. 3. As
can be seen, to fulfill the dc fault clearance, at least 50%
FBSM needs to be used for the hybrid MMC topology.
Considering the converter switch-out (from 800kV to
400kV per pole) and switch-in (from 800kV to 400kV per
pole) on-line, the VSC should have the ability to operate at
50% reduced dc voltage and 1.0p.u. dc current at the same
time. This condition makes the ratio of FBSM increase to
about 80%.
Although the hybrid MMC has the advantage of lower
Tab. 3 Performance of hybrid MMC with different ratio of FBSM
FBSM Ratio
0% 30%
Performance
DC fault clearance NO NO
Operation at 70%
NO YES
reduced dc voltage
Operation at 80%
NO YES
reduced dc voltage
Operation at 50%
reduced dc voltage with NO NO
1.0 pu dc current
Cigre Science & Engineering • N°13 February 2019
65
power loss and lower cost than the FBSM-MMC, it
destroys the modular design of MMC to some extent.
This will lead to inconvenience in terms of overhaul and
maintenance. Overall, the hybrid MMC with 80% FBSM
will be a better choice for lower power loss and lower
cost where as the FBSM-MMC will be better for more
convenient maintenance.
5. Conclusion
For the WDD HVDC, it will have the best performance
if both of the two inverters use VSC technology, and that
will also benefit the ac system. To reach ±800kV, the VSC
with double series-connected converter per pole matches
the LCC rectifier in terms of wiring, and is more flexible.
To clear the dc faults on OHL, both the FBSM-MMC and
hybrid MMC with 80% FBSMs can meet the project’s
requirements.
6. Bibliography
[1] Marquardt R. “Modular Multilevel Converter: An universal concept
for HVDC-Networks and extended DC-Bus-applications”. Power
Electronics Conference (IPEC), 2010 International, 2010.
[2] Li X, Liu W,Song Q, et al. “An enhanced MMC topology with DC
fault ride-through capability”. 39th Annual Conference of IEEE
Industrial Electronics Society, 2013
[3] R. Zeng, L. Xu, L. Yao, and D. J. Morrow “Pre-charging and DC Fault
Ride-Through of Hybrid MMC Based HVDC Systems”, IEEE Trans.
Power Delivery, vol. 30, no. 3, June 2015, pp. 1298-1306.
50% 80% 100%
YES YES YES
YES YES YES
YES YES YES
NO YES YES
Power swing blocking and tripping -
brazilian study case
R. O. FERNANDES*, T. M.T DE S. ALVES, D. B. DE OLIVEIRA,
A. C. DA R. DUARTE, A. G. MASSAUD,
Operador Nacional do Sistema Elétrico - ONS BRAZIL
Summary
The purpose of this work is to highlight the philosophy
pertaining to power swing protection as that adopted by the
Brazilian National Grid Operator (ONS). To achieve this
outcome, two large-scale oscillation protection operations
will be presented in detail, these which actually occurred
within the Brazilian Inerconnected Power System (BIPS).
The first BIPS disturbance occurred on February, 14th,
2016, at 2:00 p.m. (Brazil local time, UTC – 3h), situated
in the state of Mato Grosso, in northern region of Brazil,
where double contingencies caused a loss of synchronism
to the generating units, this due to configuration,
conditional operations and the severity of contingencies.
As a result of this disturbance, synchronization losses
occurred to the hydroelectric plants Teles Pires and
Dardanelos and a complex of other hydroelectric plants
connected to the Parecis substation. With the formation
of three simultaneous electrical centers within the BIPS,
what was then observed under this condition were the
following : the performance protection of distance in
zones 1; tele- protection schemes and TOR (Trip on
Reclose) functions associated to automatic reclosing
logic; as well as the non-opening of transmission lines,
which is a function of the oscillation blocking logic.
The second disturbance within BIPS occurred on August,
29th, 2017, at 3:42 pm (Brazil local time, UTC - 3h),
situated the northern region of Brazil, being also in the
state of Mato Grosso, which resulted in power oscillation.
In this disturbance, real PMU data was collected from the
ONS Phasor Data Concentrator (PDC) and the results
* rafael.fernandes@ons.org.br
KEYWORDS
Power Swing, Distance Protection, PMU, Real Time, Disturbance, Special Protection Scheme.
Cigre Science & Engineering • N°13 February 2019
66
proved satisfactory, via predictive analysis with the
PMU (Phasor Measurement Unit), where it is possible
to observe and then mitigate the applicability of blocking
and oscillation protection functions within the BIPS.
Several indirect methods of detecting power oscillation
are presented in the literature, however, there are still
only a few who apply phasor synchronized measurement.
The use of PMU for the detection in real time of power
oscillation is an industry trend and its use can now be
performed as a predictive measure. This work applies
synchrophasor measurement data and the results support
the feasibility of use, as a predictive tool, as a proposal for
the detection and tripping by power oscillation.
1. Introduction
The power swing is a transient event, which is primarily
caused by phenomena occurring in the electrical
system such as short circuit, line switching, generator
disconnection, loss and/or entry of large blocks of sudden
loads. These events can cause large swings of active and
reactive power, voltage loss, voltage instability and/or
instability angle between generators and consumers [1].
The system’s ability to maintain synchronism during
disturbances depends primarily on the speed of fault
elimination and the angle that exists between the voltage
sources in the pre- fault condition. The speed of fault
elimination is directly associated with the performance
of equipment protection systems. However, angular
displacement across the system depends on power being
transferred by the transmission line and the system
impedance, and conditions subject to variation of change
in the operating condition within the interconnected
system.
By opening selected points of the power system, which
establishes equilibrium between generation and load in
the sub-systems, it is possible to maintain synchronism
in each group of generators in isolated systems. As such,
the basic philosophy of power oscillation protection is to
prevent the disconnection of transmission lines during the
occurrence of stable oscillations of power and to promote
the separation of system parts during conditions of loss
of synchronism, as characterized by unstable oscillation.
Depending on the severity of the oscillation, what can then
occur is impedance migration into the operating region
of distance protection, thus causing a potential improper
operation, which could eventually lead to a cascade
resulting in a possible blackout. Therefore, protective
relays will have means of detecting synchronous (stable)
and asynchronous (unstable) phenomena and then action
appropriate decisions. When detecting synchronous
oscillation, the relay should by rights block protection
trip-away and when a power swing fails to result in a
new point of acceptable operation, the oscillation is then
considered as unstable. In such a case, this will result in
loss of synchronization between systems or groups of
generators and an issued trip, being a corrective action, is
recommended to enhance the islanding system [1,5].
According to [1] the apparent impedance as seen by the
distance relay (ANSI 21), during a power oscillation,
this tends to migrate into the operational regions of the
distance relay, which may lead to irregular performance.
In order to prevent such conditions, the power swing
blocking (ANSI 68) are inserted into the operational logic
of the distance relay so that the protection acts solely for
the short circuit conditions and remains inoperative in the
case of power oscillations. The occurrence however, of a
short-circuit during power oscillation must be detected at
the time of fault [1]. By releasing the relay from function
68, the fault should be eliminated. To differentiate such a
situation the rate of variation of the vector impedance can
be applied.
Due to the importance of power oscillations in SPS
(Special Protection Schemes), much endevour over the
years has been developed in order to ensure the correct
differentiation between short circuit situations and those
Cigre Science & Engineering • N°13 February 2019
67
typical of power oscillations. The conventional methods,
such as Concentric Characteristics and Double Blinder
[6], stand out. In [8] the criterion of the rate of variation of
the apparent impedance is incorporated.
Other non-conventional methods have been developed,
such as the Impedance Continuous Calculation Method
and Voltage Magnitude Variation Rate in the Electrical
Center [8].
In relation to BIPS, various Brazilian Agents adopt the
philosophy of blocking all the protection zones of the
distance function, and the oscillation blocking function
(68 OSB) is then enabled only in the Interconnection
circuits, with adjustments and recommendations provided
by ONS. This practice presented positive results, when
being observed in several disturbances within the system,
being that, in unstable condition, the opening of the
system at the planned points is where the oscillation trip
(68 OST) functions were enabled, thus forming stable
subsystems and minimizing the area of disturbance, as
well as load cuts.
However, with the growth of the system and the consequent
increase in its complexity, disturbances have occurred
that caused severe oscillations of power in transmission
lines, where the trip function was not enabled, and where
protection was blocked by blocking functions that could
well have been opened, thus benefiting the System and
minimizing the consequences of disturbance [4].
Yet due to the numerous configurations that factor in
multiple contingencies and circuit disconnections outside
of ONS’s study criteria, it would be impracticable to
determine with any precision, the most suitable locations of
new installation points for protection with tripping function
that might cover all possibility to guarantee stability.
Therefore, ONS began to recommend, via pre-
operational studies and by analysis of disturbances, the
non-enablement of the oscillation blocking functions,
as in this way, distance protections would be able to
act whenever there is loss of synchronism within the
System, commanding the disconnection of these lines,
thus avoiding disturbance propagation. Contrary to this,
because there are tele-protection programs associated
with distance protection zones, the following can all occur
during oscillation: variations of adjustment and differing
operational characteristics between circuits and the use
 Figure 1 - Impedance Diagram Figure 2 - Impedance Trajectory
of quadrilateral characteristics for phase-to-phase lag; the
undesired openings or even the non-opening of circuits.
All of this, should the distance units not be blocked, as
was observed in several disturbances.
2. Theoretical formulation
Figure 1 shows the voltage diagram of a charged
transmission line. The connected networks are represented
by two equivalent sources E1 and E2, respectively. The
impedances of the sources Zs1 and Zs2 correspond to the
short-circuit power of the said two sources. The angle δ
is referenced as the power transmission angle. Should
the active power transfer increase, this angle becomes
larger. The network normally operates at voltage angles
not greater than 60° to ensure its stability. If all voltages
are divided by the load current, an impedance diagram
of the line under load is obtained. The representation in
the impedance plane is performed such that the applied
distance relay is located at the origin of the RX coordinate
system, in accordance to Figure 2. With this representation,
the load impedance measured by the relay is evident and
Cigre Science & Engineering • N°13 February 2019
68
the distance to the relay’s characteristic can be determined.
As the load changes, the load impedance moves along the
trajectory as shown in the Figure (circular path). Should
E1 = E2, which can be used as a first approximation,
the impedance follows the line direction, which is
perpendicular to the sum of the impedances: Ztotal = Zs1
+ Zs2 + Z1. For the maximum active power permitted
in steady state condition δ = 90º, the load impedance
should still maintain a safe margin for the largest range
of the relay characteristic. A margin of 20% is sufficient
to prevent relay startup during load variations occurring
during normal operating conditions [1].
Hence, the diagram with the electric center and the
incursion of the impedance vector can be defined. As
the phenomenon affects all three phases simultaneously,
only the positive sequence circuit is factored for analysis.
According to Figure 1, thus:
Therefore, by varying the values of |n | and δ the
geometry of the impedance seen by the distance relay
during power oscillation is obtained.
 Figure 3 - Electric Center Representation Figure4 - Electric Center Figure 5 - R-X Plane

Considering point M as being the electric center, the
variation of the impedance moving during the oscillation
of power, in the frequency of the swing, is:
Under normal condition, the protection relays detect
oscillations up to 10 Hz. Knowing the value of the
oscillation impedance variation, obtained through the
simulation of electromechanical transient phenomena,
it becomes possible to verify the sensitivity of various
protection relays, worse case scenario for example,
factoring: δ = 180 °.
cover the more critical contingency conditions, the
presence of power oscillation or loss of synchronism can
cause the relay to perform unduly. Power oscillation is
always a symmetric phenomenon and any asymmetric
current or voltage can be adopted to release the distance
protection function.
When a single-pole-tripping occurs in applications
with automatic reclosing, the power swing is likely to
be sensitized during the single-pole dead time. It is of
the utmost importance that there is adequate oscillation
detection during these conditions. The method based on
the calculation of the continuous impedance is provided
to solve this problem [1].

2.1 Methods of impedance calculation used by
protective relays for detection of power oscillation
A sophisticated network study is necessary to determine
the correct settings for the concentric feature or the
blinders [1,4,7,10]. The settings in the relay are fixed
and will not be changed under a system change. If the
study is not properly detailed to factor in the need to
According to Figure 2, when the impedance vector
maintains a path with a tendency to completely overcome
the impedance plane delimited by the polygon, an
unstable oscillation condition is set up. When the
impedance vector characterizes a tendency to return
to the operating point without completely overcoming
the internal and external blinders, a stable oscillation
condition is established.

Figure 6 - System Impedance Equivalent Figure 7 - Impedance Diagram

Cigre Science & Engineering • N°13 February 2019

69
For the system under study in Figure 6, 1.22/9.53
represents: Z+ = 1.22 + j9.53 % and 0.32/6.1 represents :
Z0 = 0.32 + j6.1 %. The synchronization loss protection
is inserted in the protection relay installed in circuit 3
between the Substations of Ji-Paraná and Pimenta Bueno.
For this relay, using the current split factor, the impedance
as seen by the relay will be: |Ztotal| = 341.41 Ohms. The
ONS uses for the resistive and reactive adjustment of
the blinders, in primary values, as below, considering
Vbase = 230kV and MVAbase = 100.
1) External Blinder: Resistive Part=40% or 212 Ohms
and Reactive Part=50% or 265 Ohms
2) Internal Blinder: Resistive Part=30% or 159 Ohms
and Reactive Part=40% or 212 Ohms
Thus, the adjustment of the angle that will be seen by the
relay, for these adjustment conditions will be:
External Blinder
According to [1], the impedance change rate can be
calculated considering a typical oscillation frequency
and / or known, for example 2Hz, as:
1) With 3 circuits,
2) With 1 circuit,
Note that in addition to changes in the equivalents,
the topology change will also modify the value of the
impedance variation as seen by the protection relay.
Thus, because current protection relays maintain settings
for fixed power oscillation protection [1,2,10], there is a
problem in the correct determination of this event mainly
due to systemic conditions. Thus, new approaches
are gaining field of activity in relation to applications
with a trend towards real-time decision making, and in
this work the applying of synchrophasor measurement
technology will be approached as a predictive measure.
Cigre Science & Engineering • N°13 February 2019
70
2.2 The CC-PMS project (control center-phasor
measurement system)
Currently ONS is in the process of implementing its
Control Center Phaosr Measurement System (CC-PMS)
project [3]. The project will initially possess 181 PMUs,
located in 31 Substations, and the project should have
scalability for up to 1000 PMUs, covering the main
interconnections, the main transformations and the
main transmission lines. Prior to the approval of this
innovative project, ONS used - via an agreement with a
University in Brazil, PMU - data installed in all Brazilian
States. This project, known as a pilot project, provided
much knowledge and experience and served as support
for understanding during the design and implementation
phases of the CC-PMS project.
Synchrophasor measuring technology is the starting
point of an important transition in the manner in which
the system operates, and its worldwide application has
grown to a level of maturity that propels its incorporation
into the process of real-time decision making [3]. An
important result will be presented in this article as a
proposal to the use of PMU for the detection of power
oscillation, in both stable and unstable conditions.
3. Presentation of occurrences at
bips involving power oscillation
Two occurrences will be presented in this section of the
work, both of great significance within the Brazilian
electrical system involving the performance of the
68 OST function. The first one, from February, 14th,
2016, records the formation of three simultaneous
electrical centers. The second from August, 29th, 2017,
will briefly present the disturbance that occurred in
the northern area of Brazil (Acre / Rondônia) and the
Madeira River Complex, but the focus however, will
be on the use of synchronized phasor measurement in
a preventive mode, where the implemented algorithm
will be applied, verifying the slip and acceleration of the
impedance vector during the disturbance [5].
3.1. The occurence registered on february 02, 2016,
with formation of three simultaneous electric centers
With this disturbance, the synchronization losses to the
Teles Pires, Dardanelos hydroelectric plants and of the
small hydroelectric power plant complex connected in
 Figure 8 - Diagram - Area of Mato Grosso
the SE Parecis occurred, are shown in Figure 8 below,
with the formation of three simultaneous electric centers
in the BIPS. This details the protection of distance zones
in zones 1, of teleprotection schemes, of TOR functions
associated with automatic reclosing logic, as well as of
non- opening of transmission lines as a function of the
pickup of the oscillation blocking logic.
This disturbance began with a single-phase short-circuit in
the TL 230 kV Nobres - Nova Mutum C2. The fault was
located 3.3 km from SE Nova Mutum. The short circuit
was correctly eliminated in 57 ms by action of the main
and backup distance protection in the first zones in the
Nova Mutum terminal and, in zones associated with the
teleprotection schemes, and by direct trip transfer (DTT)
reception on both terminals. During the elimination of the
fault, the automatic switch-off of the TL 230 kV Nova
Mutum - Nobres C1, solely at the Nova Mutum terminal,
suffered an incorrect reception of DTT from SE Nobres,
79 ms after fault initiation – this caused by a failure in the
connection of analogue teleprotection equipment in the
SE Nobres.
After the TL 230kV Nobres-Nova Mutum, the established
configuration culminated in a power oscillation
throughout this region, with the loss of synchronism of
the hydroelectric plants of Teles Pires, Dardanelos and the
connected plants in the substation in Parecis. These plants
Figure 9 - Oscillography and impedance seen by TL Brasnorte N.Mutum C1
Cigre Science & Engineering • N°13 February 2019
71
remained out of synchronism for a few seconds until their
BIPS disconnections.
3.1.1. Registered oscillography for the three electric
centers
In the oscillography below, it is possible to observe an
oscillatory phenomenon that commenced soon after a
double circuit loss, causing loss of synchronism to the
Dardanelos hydroelectric plant with BIPS, with the
formation of an electric center in the middle of the line, as
observed in the impedance diagram.
During the oscillation, the opening of this line failed to
occur, as the oscillation blocking function (68 OSB) was
activated, preventing the operation and pickup of the line
protection zones. This line did not have trip protection due
to the loss of synchronism.
In the TL 230 kV Juína - Dardanelos C1, an oscillatory
phenomenon was also observed, immediately after the
double circuit loss, with the formation of an electric center in
the line, near the Juina busbar. After 1650 ms from the start
of the disturbance, the automatic shutdown of the TLs 230
kV Juína - Dardanelos C1 and C2 in both terminals occurred,
due to the action of the distance protections in zones 1,
where there are also no trip protection due to enabled loss
of synchronism, this being a consequence from the loss of
synchronism at the Dardanelos HPP with the BIPS.
 Figure 10 - Oscillography and impedance seen by TL Juína - Dardanelos C1
Figure 11 - Oscillography and impedance seen by TL Parecis - Brasnorte
From the double loss of circuits between Nobres and
Nova Mutum, this culminated in the loss of synchronism
of the generating units of connected hydroelectric plants,
through a transformation 138/230 kV, in the SE 230 kV
Parecis. In the TL 230 kV Parecis - Brasnorte C2, the
oscillation commenced immediately after the double
circuit loss. A change in the oscillation frequency and
trajectory as seen by the relays at the Parecis terminal
was observed after the opening of the TL 138 kV Parecis-
Telegraph, which occurred after approximately 1633
ms from the beginning of the disturbance. Prior to the
opening of the 138 kV circuit, the blocking function was
activated by oscillation, thus preventing the operation of
the protection zones. It should be noted that, in this case,
there was no formation of an electric center in line.
The oscillatory process finally terminated after the
disconnection of the Teles Pires hydroelectric plant -
which were out of synchronism, due to over-frequency
protection. After the Teles Pires power plant was shut
down, the oscillatory process decreased in the region,
however, as a consequence of generation loss in the
area, there were over-voltages in the 230 kV circuits in
the region, resulting in the automatic shutdown of the
following circuits in function of overvoltage (ANSI
59T) of TLs in 230 kV Brasnorte – Parecis C1; Nova
Cigre Science & Engineering • N°13 February 2019
72
Mutum – Sorriso C2 and Brasnorte – Nova Mutum
C1 e C2.
3.2. Registered occurence in August 29th, 2017,
involving the northern region of Brazil
This disturbance commenced at 15h:48min, (Brazil
local time, UTC-3h) with the occurrence of a two-phase
short circuit, involving phases A and B of TL Porto
Velho - Abunã (C1), eliminating the performance of its
protections at both terminals. Following the shutdown
of the TL 230 kV Porto Velho - Abunã C1, there was a
period of under-voltage with slow voltage recovery in
the areas of Rio Branco and Porto Velho, causing a loss
of load in Porto Velho, at those substations indicated in
Figure 12 below. After a further 10 seconds, a new load
loss occurred, now in the Rio Branco region, causing
power oscillation culminating in the disconnection of
the 230 kV TL Ji-Paraná - Pimenta Bueno (C1 / C2 / C3
) by operation of the out-of-step protection (68 OST),
opening the interconnection and provoking the islanding
of the Acre (SE Rio Branco I) / Rondônia (SE P.Velhlo)
system with the Samuel hydroelectric plant and the
“Back-to-Back” of the Madeira´s River HVDC system.
The amount of interrupted load in the states of Acre and
Rondonia area totalled 711 MW.
 Figure 12 - Diagrams of Areas Involved with the Disturbance

3.2.1. Offline simulation

ONS, via its special electrical studies division, performed
offline simulations for the correct adjustment of the power
oscillation trip function (68 OST) in these transmission
lines in the region of Mato Grosso. The result obtained,
according to Figure 13, is in agreement with that expected
and that registered by the PMU. The adopted software for
electromechanical simulation was Anatem / Cepel and the
database was that of 2016, with the same pre-fault operating
point recorded by ONS’s state estimator.
This simulation showed that the power oscillation was
dampening and the system would tend to return to a stable
operating point. However, the actual 40-MW loss in the Rio
Branco system was not simulated, which in practice has led
the system to instability, as can be seen in Figure 14 below,
this from the actual records obtained by the PMUs.
3.2.2. Registration from phasor measurement unit (PMU)
With a PMU recording, at a rate of 60 frames / sec, its
possible to check the phenomena recorded in the Figure
below, however, it was observed that the system was driven
to an unstable operating point (formation of an electric
center on the transmission line Ji Paraná - Pimenta Bueno,
in 230 kV) due to the loss of load in Rio Branco, culminating
in the performance of trip protection by power oscillation.

Figure 13 - Simulated System without Load Rejection - Power Swing Stable

Figure 14 - PMU Measurement during Disturbance

Cigre Science & Engineering • N°13 February 2019

73
 Figure 15 - Logic Diagram - Power Swing Detection

4. The use of PMU for the power
oscillation detection on the
brazilian interconnected power
system (BIPS)
According to the possibilities of utilizing synchronized
phasor measurement in [1-3-10] and with the results
obtained through simulations in [5], this work was
motivated to apply these measurements in order to obtain
a proposal for the Predictive power swing protection
based on PMU only.
and OST conditions. Contained in this same work,
the following logical condition of detection of power
oscillation based on PMU is presented.
In the Figure above, the threshold adjustments for
acceleration and slip were obtained according to the
system analyzed in [5], and each system analyzed will
have its own adjustment thresholds. Therefore, the
detection of the power oscillation will depend solely
on those values obtained by the PMU, with the slip and
acceleration being based purely on the angular difference
between two substations.

According to [5], the synchropasors within protective
relays have been available since 2002. Typical applications
of this technology are visualization, state measurement
and system integrity protection schemes.
By measuring the angular difference between two
substations, here represented by the Substations Ji Paraná
- Pimenta Bueno, it was possible to calculate the slip
frequency and acceleration to identify power swings
4.1 Results obtained from pmu records
According to Figure 16 it is possible to observe the pre-
fault operating point being the instant of the short circuit
(t0 +), the balance of the generation x load after the
elimination of the short circuit, the beginning of the power
oscillation due to rejection of load, and finally the instant
of time in which the impedance vector incursion to trip
area. ONS determines that should the impedance vector

Figure 16 - Diagram R-X and Impedancy Trajectory during Disturbance

Cigre Science & Engineering • N°13 February 2019

74
 Figure 17 - Angle Difference Trajectory seen by PMU during Disturbance

enter the outer blinder and then exit the first internal
blinder in less than 25 ms, a trip command due to power
oscillation must be issued. It is worth noting that the
external and internal blinder are adjusted as presented in
this work, item 2, Figure 7.
Figure 17 above shows the angular difference trajectory.
The instant of the short circuit was at 15:48min25s63ms.
The PMU time base is UTC + 3h, then 18h48min25s63ms.
Internally the PMU converts this UTC time to an internal
value known as SOC (Second of Century) which is the
total seconds since the first instant of the year 1970.
From the graphic of the Figure below, the first instant
of the short circuit is represented by time 505 s, so as to
facilitate the calculation of time difference in Figure. It
is important to note that the pre-fault angular difference
between equivalent internal sources was 29º. After
eliminating the short circuit, there is a balance between
the generation and load and the angular difference
between the equivalent sources is around 41º. As a result
of load rejection, a period of power oscillation begins,
with the angular difference varying around 60º. Finally,
the trip signal is issued by power oscillation protection
with the angular difference between sources around 88 °.
During the current elimination, with the total opening of
the circuit-breaker, the angular difference between the
sources is around 98º.
According to Figure 18 below, the slip and the angle
difference in trajectory during the disturbance are
verified. It is important to note that during the load
rejection event, which concluded in an unstable power
oscillation process, the slip, which was approximately
zero, with an angular difference between sources around
62º, tending to attain a stable operating point, jumped to
6 °/s, approximately. After this point in time, the angular
difference between the sources increased significantly, as
did the slip. At the point of the power swing, the angular
difference was approx 88 °, as previously recorded, and
the slip reached 40 °/s.

Figure 18 - Slip and Angle Difference Trajectory during Disturbance

Cigre Science & Engineering • N°13 February 2019

75
 Figure 19 - Slip and Acceleration Trajectory with Time Duration during Disturbance

In the Figure 19 above, it is possible to verify the slip
and angle difference evolution with the time. The instant
of time that the out-of-step condition is detected by the
method proposal in this paper is 524.5 s, and the instant
of time that the trip is issued by protective relay is 524.9s,
approximately 400ms of time difference.
In the Figure 20 below, the slip and acceleration values
for the said disturbance are finally verified. It is of
interest to note that at the point of load rejection, which
culminated in unstable oscillation, the acceleration was
near to 0.35 Hz/s and the slip was approx 0.048 Hz,
and with oscillation in the system no longer returning to
equilibrium. The proposed settings, for this new approach
analysis, are, for example, Acceleration = 7 Hz/s and Slip
= 0.11 Hz, and the trip characteristic slope is indicated by
red line. It is important to note that during trip by out-of-
step protection, the Acceleration reached approximately
11 Hz/s and the Slip is near to 0.11 Hz.

Figure 20 - Slip and Acceleration Trajectory during Disturbance New Approach Analysis

Cigre Science & Engineering • N°13 February 2019

76
5. Conclusions
This work highlights the performance of trip protection by
oscillation of power, culminating in the formation of three
simultaneous electrical centers in an important northern
region of Brazil. This experience in Brazil with relation
to the application of the power oscillation function was
demonstrated via two real oscillographs, where it was
possible to verify the difficulty in relation to the correct
detection of the oscillation and the threshold between
the block and the trip, primarily due to the topology
changes and unforeseen conditions during the periods of
simulation. The consequences were also exposed within
the electrical system, resulting in a blackout of sizable
magnitude.
Finally, the oscillatory process ceased after the
disconnection of the Teles Pires plants - which were
connected out of synchronism, due to the overfrequency
protection. After the Teles Pires power plant was shut
down, the oscillatory process decreased in the region,
however, as a consequence of the generation loss
throughout the area, there were overvoltages in the 230
kV circuits of the region, which resulted in the automatic
shutdown of the following circuits and timed overvoltage
protections within the electrical system.
As a result of implementing the PMU project, as overseen
by ONS, and also via new algorithms that do not require
system adjustments, such as equivalent impedance values,
the application of synchronized phasor measurement
has now become a reality, although its principal focus
still appears to be on the monitoring of electrical system
quantities and support for event analysis.
The results obtained in this work with the actual use of
PMU data, collected in ONS’s Phasor Data Concentrator,
revealed the potential of detecting power oscillation
and its triggering - under unstable conditions, as a
predictive procedure, solely via the measurement of
angular difference between sources without sophisticated
adjustments to existing protection relays. In relation to
real-time decision making, the latencies are measured
between the data externalized by the PMU until such time
as its arrival in the data concentrator when a dedicated
communication channel is utilized. This is less than 80
msec, so it is estimated that a total local actuation time can
be approx 100 msec.
Cigre Science & Engineering • N°13 February 2019
77
According to the results as seen in Figure 19, the time
difference between the detection of unstable oscillation,
by the method presented in this paper, and the detection
by protective relays, represented approximately 400 ms.
This result shows the possibility of reducing the current
time adopted by protective relays, which uses only the
impedance characteristic to issue the trip signal during the
out-of-step condition.
The WAMPAC (Wide Area Monitoring Protection
and Control) applications have become a reality in the
operation control centers of TSO (Transmission System
Operator) and continue to inspire development engineers
to apply new algorithms via synchrophasor measurement
technology.
This work highlights this viability by employing real
PMU data, where the gains obtained through the reduction
of fault clearing time elimination on the protective relays
considering the systemic conditions in the BIPS and
where there is an oscillation of unstable power can prove
to be substantial, for example, avoiding a large blackout
in the system.
6. Bibliography
[1] R.O. Fernandes, B.G. Isolani, S.A. Souza, “Techniques and Methods
of Power Swing Detection to avoid Blackouts”, in Portuguese,
Florianópolis, Brazil, XI STPC, November 18-21, 2012.
[2] R.O. Fernandes, M. E.C. Paulino, “Power Swing Secondary
Testings and Analysis of Transient and Steady State Performance of
Protective Relays”, Belo Horizonte, Brazil, Cigré Study Committee
B5 Colloquium, August 25-31, 2013.
[3] R.O.Fernandes, A.Massaud, M.Muniz, S.Diniz,H.Volskis, “The
CC-PMS Control Center Phasor Measurement System in the BIPS
(Brazilian Interconnected Power System) and the Lessons and
Learned with the PMU Data Acquisition Pilot Project through
openPDC by the ONS Power System Studies and Event Analysis
Team”, in Portuguese, Curitiba, Brazil, XXIV SNPTEE, October,
22-25, 2017.
[4] T.M.S. Alves, D.B. Oliveira, A.C.R. Duarte, A.G. Massaud, A.F.C.
Aquino, A.N. Nohara, “Application of the Blocking and Tripping
Protection Functions by Power Swing” (68 OSB and 68 OST) in
the BIPS (Brazilian Interconnected Power System) – Real Cases”, in
Portuguese, Brasilia, Brazil, XIII STPC, October, 23-27, 2016.
[5] A. Guzmán, M. Venkat, G. Zweigle, “Backup Transmission
Line Protection for Ground Faults and Power Swing Detection
Using Synchrophasors”, San Diego, United States, Distributech
Conference, February 3-5, 2009.
[6] J. Holbach, “New Out of Step Blocking Algorithm for Detecting
Fast Power Swing Frequencies”, in Power Systems Conference
2006: Advanced Metering, Protection, Control, Communication,
and Distributed Resources, pp. 182-199. [7] IEEE PSRC WG D6,
 “Power Swing and Out-of-Step Considerations on Transmission
Lines”, Jul. 2005.
[8] B. Alinejad, H.K.Karegar, “A Novel Out-of-Step Relay Using Wide
Area Measurements”, The 5th Conference on Thermal Power Plants
(IPGC2014), June, 10-11, 2014.
[9] Abidin, A.F., Mohamed, A. and Shareef, H., “Power swing detection
for correct distance relay operation using S-transform and neural
networks”, in Conference Publications 2011: Power Engineering
and Optimization Conference (PEOCO), 6-7 Jun.
[10] NERC - North American Electric Reliability Corporation
“Protection System Response to Power Swings” System Protection
and Control Subcommittee, August 2013.

Cigre Science & Engineering • N°13 February 2019

78
Planning a meshed HVDC offshore
grid in the North Seas
P. Henneaux*, M. Lessinnes, A. Mannocchi, K. Karoui, Tractebel, Belgium
J. Bettink, N. de Groot, M. de Schepper, TenneT, The Netherlands
Summary
In many regions of the world, offshore wind energy
is expected to become a major energy resource in the
upcoming decades. This is the case for the European
North Seas, where up to 205 GW of installed offshore
wind parks are expected to be developed by 2050
according to the most ambitious scenarios. To transfer this
offshore wind energy to the shore, a dedicated network
infrastructure will be required. Because a significant
part of this wind energy will also be located far from the
shores, the use of HVDC technology will probably be
required. Thanks to the ongoing development of HVDC
breaker technology and related protection systems, a
meshed HVDC offshore grid linking offshore wind
parks and onshore grids in different countries can be
developed. Several studies already demonstrated that
the development of such a grid in the North Sea can
bring substantial benefits. Such a grid can be used not
only for the evacuation of the offshore wind energy, but
also to exchange power between the different North Sea’
countries and possibly other benefits such as supporting
the stability of onshore AC grids. The purpose of this
paper is to present the results of a prefeasibility study
of a meshed offshore HVDC grid in the North Sea
by looking at both the economic and the technical
feasibility, carried out in the context of the European
H2020 project “PROMOTioN”. A prerequisite of this
prefeasibility study is the definition of a set of technical
criteria that need to be fulfilled by the meshed HVDC
offshore grid. If planning criteria are well defined for
onshore grids (e.g. N-1 security criterion), this is not the
case for offshore grids: there is no standard guideline
*pierre.henneaux@tractebel.engie.com
KEYWORDS
Power system planning, HVDC transmission, DC transmission grid, offshore, wind power, North Seas, meshed HVDC
grid
Cigre Science & Engineering • N°13 February 2019
79
to plan such a grid. This paper begins thus with the
drafting of a set of planning criteria for an offshore
grid, based on a review of planning criteria for onshore
grids, for connection of offshore wind farms and for
HVDC interconnectors. Furthermore, there is no unique
way to develop a European offshore grid. Several
strategies of development are possible and feasible,
which could consequently lead to very different offshore
grid structures (topologies). The challenges will be
dependent on the strategy applied. This paper develops
possible “Concepts” for a future offshore grid. Finally,
two specific concepts are analysed in more detail, by
developing the corresponding grid structure over the time
period 2020-2030, and by analysing the corresponding
costs and benefits.
1. Introduction
Offshore wind energy is a major renewable energy
resource in several regions of the world, and in particular
in the European North Seas (i.e. North Sea, English
Channel, Irish Sea, Skagerrak and Kattegat). For that
reason, numerous offshore wind farms are expected
to be commissioned in the upcoming decades in the
North Seas. The actual development pace is uncertain,
but, starting from 27 GW in 2020, between 38 GW
and 65 GW of installed capacity is expected in 2030,
and between 90 GW and 205 GW in 2050. If offshore
wind farms are currently mainly connected radially and
individually to onshore grids, the HVDC Voltage Source
Converter (VSC) technology makes the development
of coordinated solutions possible, connecting several
offshore wind farms to the shores through an offshore
grid. Indeed, both radial multi-terminal and meshed
grids can be developed to create synergies between
the connections of the various offshore wind farms to
the shores. This way, it could be possible to evacuate
all the offshore wind energy at a lower cost. For
example, several studies have shown that a coordinated
development of offshore electricity infrastructure in the
North Seas can bring significant financial, technical, and
environmental benefits [1].
The development of such a meshed offshore HVDC grid
encompasses several challenges. First, if planning criteria
are well defined for onshore grids (e.g. N-1 security
criterion), this is not the case for offshore grids: there is
no standard guideline (e.g. requirements on the system
performance under normal and contingency conditions)
to plan such a grid. Second, there is no unique way to
develop a European offshore grid. Several strategies
of development are possible and feasible, which could
consequently lead to very different offshore grid
structures (topologies). These strategies require diverse
levels of coordination between stakeholders and lead to
various challenges. Third, precisely because numerous
offshore wind farms are expected in the upcoming
decades, and because multiple onshore connection points
are possible, there is a combinatorial complexity in the
choice of cables to install to evacuate the offshore wind
while satisfying the planning criteria. Fourth, the HVDC
technology imposes specific constraints that must be
considered in the planning phase, such as the voltage
level at which to operate. Consequently, the planning of
coordinated connections of offshore wind farms to form
a meshed HVDC offshore grid is much more challenging
that the planning of individual connections.
The purpose of this paper is to establish the first steps
needed to plan a meshed HVDC offshore grid in the
North Seas and to present the results of a prefeasibility
study of a such a grid by looking at both the economic
and the technical feasibility. It corresponds to the
preliminary analysis performed by the European H2020
project “PROMOTioN”. Because there is uncertainty on
costs and capabilities of components, numerical results
must be taken with care. This paper is thus organized as
follows. Section
2
Note that the need to use probabilistic criteria (i.e. a risk of problems is acceptable but must be below a given threshold) to complement (or replace) deterministic criteria
is rising.
Cigre Science & Engineering • N°13 February 2019
80
2 drafts a set of planning criteria for an offshore grid,
based on a review of planning criteria for onshore grids,
for connection of offshore wind farms and for HVDC
interconnectors. Section 3 develops possible
“Concepts” for a future offshore grid. Section 4 analyses
then a specific concept more in detail, by developing
the corresponding grid structure over the time period
2020-2030, and by analysing the corresponding costs
and benefits. Finally, section 5 concludes and gives
perspectives for further work.
2. Planning criteria for an
offshore grid
Planning a transmission power system consists in
finding the most economical way to develop the system
while being able to satisfy components operational
limits and maintaining an acceptable reliability level,
from both an adequacy and a security point of view. The
set of constraints that must be fulfilled by the planned
power transmission system defines the planning criteria.
Transmission planning criteria typically cover (but
not limited to): system states and the contingencies
(including faults) to study, acceptable system operating
limits in normal operation (pre-contingency) and post-
contingency states, acceptable response of the system to
outages and to fault disturbances. The security analysis
is traditionally based on deterministic criteria2 (i.e.
the system must ride through the postulated secured
contingencies while staying within acceptable operating
conditions). Although planning criteria for onshore grids
are not fully the same in every country/region of the
world, a common approach is usually followed, and it is
thus possible to deduce “typical planning criteria”. This
is not the case for meshed offshore HVDC grids: no
planning criterion formally exists for that kind of grid.
The purpose of this section is to provide a preliminary
discussion on potential planning criteria for an offshore
grid. It is thus organized as follows. Subsection 2.1
reviews existing planning criteria. Subsection 2.2
analyses planning criteria that are currently used in
Europe to plan HVDC offshore wind farms connections
and existing HVDC interconnectors. Finally, subsection
2.2 discusses a first proposal of planning criteria for
offshore HVDC grids that could be applied to purely
radial, radial multi-terminal and meshed structures.
2.1. Review of planning criteria for an onshore grid
For onshore grids, conventional transmission planning
criteria have been established at national, regional or
utility level. Although they differ in implementation
aspects, the overall philosophy is the same. The planning
criteria have first to ensure that the transmission power
system can accommodate the load and the generation
under normal conditions while satisfying operational
limits and being stable. In planning, the system is said
to be under “normal conditions” typically when all
generation and transmission facilities are available. To
cover the possible range of operating conditions, system
peak load and system off-peak load are usually analysed.
The massive integration of renewable energy sources
leads nevertheless to critical operating conditions
not covered by these two conditions. Therefore,
other critical conditions can also be analysed. For all
conditions studied, the power transmission system must
accommodate the power flows given by the economic
dispatch (or another credible dispatch) of generating
units, without load shedding, with power flows through
transmission elements within normal (continuous)
ratings and voltages at nodes within specific. The
planned power system must also be such that the
maximum short-circuit currents are below the ratings of
circuit breakers. Planning criteria include also typically
constraints linked to the system performance following a
single contingency, the so-called “N-1 security criterion”.
A single contingency can be defined as the trip of a
single network element (e.g. single transmission circuit,
generator, reactive compensator) that cannot be predicted
in advance. Transmission systems are typically planned
such that a single contingency occurring at system peak
or off-peak load does not endanger the system security:
the system must stay electrically, loss of load (i.e. load
shedding) is not allowed, no uncontrolled cascading
outage is allowed, and power flows and voltages
must be within specific operating limits just after the
contingency. Planning criteria detail usually the types of
contingencies to consider, the types of faults to analyse in
the stability analysis, the method to perform the stability
analysis, the allowed operating limits, etc. If the analysis
Cigre Science & Engineering • N°13 February 2019
81
of the system performances under normal conditions and
following a single contingency are standard procedures
during the planning stage, this is not the case for the
analysis of the system performance following several
or extreme contingencies. Nevertheless, it is included
in planning criteria for various countries, with different
requirements. The general purpose is to check that
contingencies more severe than the ones covered by
the N-1 security criterion, but nevertheless with a
non-negligible likelihood, do not entail unacceptable
consequences.
2.2. Analysis of current planning criteria for
European HVDC systems
Because HVDC offshore wind farm connections are
part of the transmission system, the philosophy behind
planning criteria stays the same as that for onshore
grids: any HVDC offshore windfarm connection must
accommodate the generation under normal conditions
(i.e. when all transmission facilities of the connection
are available) while satisfying operational limits and
being stable, and a single contingency cannot endanger
the system security of the onshore grid (N-1 security).
However, as the HVDC offshore windfarm connections
connect almost exclusively generation to onshore grids,
and they do not connect offshore loads (except for the
auxiliary equipment that consumes electrical power),
implementation details differ. In particular, it is not
relevant to perform the analysis for peak and off- peak
load conditions. In the case of an offshore wind farm
connection, the analysis is done at the nominal capacity
of the wind farm (i.e. peak generation). Note that the
N-1 security criterion for the connection of an offshore
wind farm does not mean that the generated offshore
wind energy must still be evacuated after a single
contingency. It means that a single contingency cannot
lead to an unacceptable disturbance in the onshore grid,
such as a load shedding.
The European network code on requirements for grid
connection of high voltage direct current systems
and direct current-connected power park modules [2]
establishes general requirements on planning criteria
for HVDC systems (e.g. HVDC offshore wind farm
connections and HVDC interconnectors between
different synchronous areas). Article 17§1 could be
interpreted as follows: following a single contingency
within the HVDC system, the loss of active power
injection in a synchronous area shall be limited to a
value specified by the relevant TSOs for their respective
load frequency control area.
2.3. Proposition of planning criteria for an offshore
grid
The planning criteria of an offshore grid must first
ensure that the offshore grid can accommodate the
offshore load and the offshore generation under normal
conditions while satisfying operational limits and being
stable. It is proposed to consider, in planning, that the
offshore grid is under “normal conditions” when all
converters and transmission facilities are available
(similarly to onshore grid). It is then necessary to define
the specific operating conditions that must be studied at
the planning stage. These selected operating conditions
should somehow cover all the possible ranges of
operating conditions of the offshore grid. Because the
offshore peak load (e.g. offshore oil/gas platforms) is
expected to be much lower than the installed offshore
generating capacity, the peak load is not expected
to be a critical condition. On the contrary, the peak
generation is expected to be a critical condition. It is
thus proposed to analyse the peak generation, when
all offshore wind generators produce at their nominal
rating (i.e. maximum power output). Contrary to the
AC grids, the off-peak generation (i.e. when no offshore
wind generator produces power) is not expected to
be a critical condition in an HVDC grid. Therefore,
these provisional planning criteria propose to limit the
analysis to the offshore peak generation. In the future
it could also be considered to optimize further and
decrease the connection capacity. Note that, in a radial
multi-terminal or a meshed grid, the generation level
does not lead uniquely to the power flows in the grid:
the set-points of onshore converters also impact these
power flows. Finally, the operational limits must be
defined. For AC grids, there are two sets of operational
limits considered in the planning stage in steady-state
analysis: thermal limits and voltage limits. It is proposed
to have the same for offshore HVDC grids. The
requirement is straightforward for power flows through
transmission elements: under normal conditions, power
flows must be below the normal (continuous) rating
(no overload). For voltages at nodes, the requirement
Cigre Science & Engineering • N°13 February 2019
82
is more complex, because there is no standard value for
acceptable voltages within an HVDC grid. In line with
what is done for onshore grids, it is proposed to require
as a provisional planning criterion that, under normal
conditions, voltages at all nodes must be between 0.95
pu and 1.05 pu. The proposed planning criteria for the
system performance under normal conditions can be
summarized as follows. When all HVDC transmission
elements are available and all offshore wind generators
are at their maximum (nominal) power output, it must
be possible to set the onshore converters such that
power flows within HVDC transmission elements are
below the normal (continuous) rating and that voltages
at all nodes of the offshore HVDC grid are between 0.95
pu and 1.05 pu. This does not yet consider the country
location of the windfarm and possible constraints: wind
produced in country A does not always have to be able
to be transported to country A to fulfil the criterion. The
concepts below describe different pathways where these
regulatory approaches are part of the planning criteria.
The pre-contingency state to consider for the analysis of
the system performance following a single contingency
is the one considered for the analysis of the system
performance under normal conditions: the peak
generation (i.e. all offshore wind generators are at their
maximum power output) with all HVDC transmission
elements available. It is proposed to consider as single
contingency the loss of a converter, the loss of a cable
with or without a fault and the loss of an overhead
line (OHL) with or without a fault. In line with what
is done for AC grids, the requirements on the system
performance following a single (N-1) contingency are
proposed to be the following: (i) the system must stay
electrically stable; (ii) no uncontrolled cascading outage
is allowed (but the disconnection of an offshore wind
farm radially connected, or an action of an automatic
Remedial Action Scheme is allowed); (iii) power flows
and voltages must be within emergency operating limits
just after the contingency, once the automatic voltage
droops of converter controller have stabilized the
system, and they should go back to normal (continuous)
operating limits after system adjustments; (iv) the
permanent losses of power infeed into an onshore
grid must be below the reference incident of the
corresponding synchronous area (e.g. currently 3000
MW for continental Europe).
 Figure 1. Business-as-usual concept.
3. Concepts
A future European offshore grid can be developed
following different strategies that are called “Concepts”
in this paper. These strategies require diverse levels of
coordination between stakeholders (e.g. countries, TSOs,
offshore wind developers), face various challenges,
and could consequently lead to very different offshore
grid structures (topologies) and governance. Although
it does not lead to a meshed offshore grid, the current
way the grid is developed in a weakly coordinated
way, i.e. individual connections of offshore wind farms
decoupled from interconnectors, can be considered as a
first possible strategy to develop offshore infrastructure
and can thus be called “Business-as-usual concept”.
A first possibility to develop a grid in a coordinated
way corresponds to the “North Sea Wind Power Hub”
proposed by TenneT and Energinet in 2016. In this case,
offshore wind farms are connected to large centralised
connection points, and power is evacuated to North Seas
countries through dedicated HVDC corridors. A second
option could be the creation of small HVDC hubs, like
the current AC substations in the onshore grid, meshed
at the national level but loosely interconnected between
different countries. Lastly, a third option could foresee a
stronger interconnection between decentralized hubs up
to a level where the overall infrastructure forms a fully
meshed international offshore grid.
The following subsections describe these different
concepts. The Concepts are sharply contrasted to make
them distinctive, but they are not incompatible. Indeed,
it is expected that a future offshore grid will be the
combination of different concepts. In particular, for the
sake of readability, radial connection of offshore wind
farms and offshore point-to-point interconnectors are
not shown in the advanced concepts, even while they
will remain an important part of the picture.
Cigre Science & Engineering • N°13 February 2019
83
3.1. Business-as-usual concept
Before delving into the concepts of offshore grids, it is
worth considering a first possible development of the
offshore infrastructure in the North Sea in a “Business-
as-usual concept”, i.e. if no action is taken to develop an
offshore grid in a coordinated way. This business-as-usual
concept can be used as a reference case to benchmark
the other concepts. In this concept, each country
develops its own wind generation connected radially to
the mainland with the sole purpose of evacuating wind
generation to its own power system. Offshore wind
farms are connected either individually to the shores,
or via small hubs. However, offshore hubs connect only
the offshore generation of the same country. In parallel
to the evacuation of the offshore wind power, point-to-
point interconnectors are developed to exchange energy
between countries. Figure 2 below shows an example of
this business-as-usual concept.
3.2. Centralized wind power hubs concept
The concept of centralised wind power hubs describes
a potential future development of the offshore grid
polarised around AC hubs to which several wind farms
are connected, in line with the “North Sea Wind Power
Hub” proposed by TenneT and Energinet. Figure 2
provides an example of how such a grid could look like.
Offshore wind farms are connected radially to the AC
hubs, mainly through AC connections, but potentially
through DC connections when the distance justifies it.
The AC hubs are connected directly to the mainland of
one or more countries through point-to-point HVDC
corridors for the evacuation of bulk wind generation.
Eventually, these hubs could be interconnected with
each other, providing both additional evacuation routes
for the offshore wind farms and interconnection capacity
between different market areas. Aside from the central
hubs, offshore wind farms close to shore are connected
 Figure 2. Centralized wind power hubs concept.

radially to the mainland, or through small hubs, as per the
respective national policies. It should be noted however,
that due to the technical challenges involved building
the artificial infrastructure (islands), this a rather long-
term approach which is not likely to materialize prior to
2030/35, according to its promotors.
3.3. National decentralized wind power hubs
concept
In contrast with large centralized wind power hubs, the
offshore grid could be developed around several small
hubs, interconnected amongst themselves to different
degrees. Such approach can branch off in two specific
concepts for wind power hubs, based on a national or
European approach to policies for the offshore grid.
The national decentralized wind power hubs concept,
illustrated in Figure 3, achieves an offshore grid based
on distributed wind power hubs by promoting the
development of meshed national offshore grids, weakly
interconnected amongst each other. In this concept,
each country develops its own national offshore grid
according to the national policies. The scope of the
offshore grid is to evacuate all the national offshore wind
power generation to the corresponding onshore national
grid. With respect to the radial connections of wind
farms, a meshed offshore grid grants more flexibility
and increased security as wind power can be injected
in the mainland system through different connection
points depending on the state of the system. The national
offshore grids might be interconnected with each other.
The envisaged “Wind connector” Concept between the
Dutch area “IJmuiden Ver” and the British “East Anglia”
area corresponds to this Concept: the Dutch offshore
grid would have to be designed to evacuate all the Dutch
offshore wind energy to the Dutch onshore grid, the
British offshore grid will be designed to evacuate all
the British offshore wind energy to the British onshore
grid, but the two offshore grids would be connected by
an interconnector.

Figure 3. National decentralized wind power hubs concept.

Cigre Science & Engineering • N°13 February 2019

84
 Figure 4. European decentralized wind power hubs concept.
3.4. European decentralized wind power hubs
concept
On the other hand, the European distributed wind
power hubs concept, illustrated in Figure 4, favours
the joint development of a single MOG integrating all
the distributed hubs. In this concept, the distributed
wind power hubs are integrated into a strongly meshed
offshore grid developed jointly by the North Sea
countries as an international infrastructure. Offshore
wind farms can be connected either in AC or DC to the
most suitable node of the grid regardless of where it
belongs to. The onshore grid is not anymore designed to
be able to evacuate national offshore wind energy to the
corresponding national offshore grid. It is worth noting
that the national decentralized wind power hubs can
eventually evolve into the European one by increasing
the integration between the national offshore grids.
4. Examples
This section aims at analysing more in details two of the
proposed concepts, the business-as-usual concept and
the European decentralized wind power hubs concept,
by developing corresponding offshore grid topologies
for the period 2020-2030. The detailed grid structures
(i.e. converters, cables, substations, circuit breakers)
for these two development frameworks satisfying the
planning criteria are optimized. The technical viability
of the obtained structures is then checked based on
power flow analysis under normal conditions. Finally,
the economic viability of the two network development
plans is analysed through market simulations. This
section is organized as follows. Subsection 4.1 details
the methodology. Subsection 4.2 summarizes the key
assumptions. Subsection 4.3 presents the results. Detailed
explanations about the methodology, the assumptions
and the results can be found in the PROMOTioN’s
“Deliverable 1.6: Draft roadmap and reference offshore
Cigre Science & Engineering • N°13 February 2019
85
grid expansion plan” [3]. Note that these are the outcomes
of a prefeasibility study, which aimed at defining
the key factors influencing the planning of a meshed
offshore DC grid. Not the topologies and the locations
of infrastructure are of interest, but the structure of the
proposed topologies that were suggested by the model.
4.1. Methodology
The approach followed to develop grid topologies for
each concept follows three steps. The first step aims
at determining the transmission needs to evacuate the
offshore wind energy. It is formulated as a multi-period
Optimal Transmission Expansion Planning (OTEP)
problem with simplification of the technical constraints
to consider. The objective function of this OTEP
problem corresponds to the offshore grid infrastructure
cost (i.e. CAPEX), actualized over the period considered
(i.e. 2020-2030). Transmission capacity between two
substations is modelled in a discrete fashion: binary
decision variables represent the possibility to install a
circuit between substations of the system. Constraints
of the OTEP problem impose to the offshore grid to
evacuate the peak generation of offshore wind farms
while satisfying operational constraints (e.g. power
flows within limits). The OTEP problem leads then to
a large-scale Mixed Integer Linear Program that can be
solved by branch-and-bound algorithms. The second
step aims then at refining the design of the offshore
grid such that the resulting grid is technically viable
and meets the planning criteria. Indeed, because some
simplifications are made for the offshore grid model in
the formulation of the OTEP problem (e.g. transportation
model), the technical viability of the proposed solution
while considering the full power flow equations must be
checked and adaptions must be made if the solution of
the OTEP problem does not comply with the planning
criteria. Solving the power flow equations for the
offshore HVDC grids is thus the first step of the technical
 Figure 5. Business-as-usual concept: example of a possible resulting grid in 2030.
analysis. Power flow equations for HVDC grids are
used for that purpose, as described in [4]. Moreover,
the optimization problem focuses on the topology and
does not consider converters and circuit breakers. This
second step determines also which converters can be
used and where DCCBs are needed. Finally, the third
step evaluates the economic viability and the level
of adequacy of generated offshore grid topologies by
assessing more precisely the benefits brought by the grid,
especially the generation cost savings corresponding to
the change in the socio-economic welfare (SEW). The
latter is estimated through a market study, to simulate
the dispatch of generating units by mimicking perfect
market behaviour. The perimeter of the market model
includes the North Sea’s countries.
4.2. Key assumptions
Offshore grid topologies are developed for the period
2020-2030. It is considered that interconnectors and
offshore wind farms connections already existing or
that will be commissioned by 2020 cannot be part of a
meshed offshore grid, because they were not designed
in that way (e.g. different voltage levels are used). The
load/generation scenario adopted corresponds to the
Vision 3 of the ENTSO-E TYNDP2016 for 2030. In the
North Sea itself, 37 GW of offshore wind is expected
to be commissioned between 2020 and 2030 with
approximately half in the United Kingdom. These global
figures can be translated in concrete projects using
the 4-C-Offshore wind database containing detailed
information on individual projects (geographical location,
target wind farm size, current development status). A
homogenous voltage level of 525 kV was chosen, in line
with the recommendations of the Cigré Joint Working
Group B4/C1.65 for the range of power involved [5].
Costs of HVDC converters (VSC) and HVDC cables
are taken from publicly available information. On the
contrary, because the technical capabilities and costs of
HVDC circuit breakers (DCCBs) are not known, two set
Cigre Science & Engineering • N°13 February 2019
86
of assumptions are adopted: in one case, it is assumed
that expensive DCCBs (e.g. 30 M€ for a 1 GW DCCB)
must be used, and, in the other case, it is assumed that
cheap DCCBs (e.g. 1 M€ for a 1 GW DCCB) can be
used. In addition, it is considered that expensive DCCBs
need a specific platform when they are installed offshore
(e.g. extra cost of 20 M€ for 1 a GW DCCB). Note that
a naïve strategy is used to protect the grid: it is assumed
that DCCBs are required for each element belonging
to a part of the grid connecting more than the allowed
permanent loss of power infeed of the onshore grid.
Detailed assumptions can be found in the PROMOTioN’s
“Deliverable 1.6: Draft roadmap and reference offshore
grid expansion plan” [3].
4.3. Achieved results
Figure 5 shows the achieved results for the business-
as-usual concept, while Figure 6 shows the achieved
results for the European decentralized wind power
hubs concept. These results are draft because there
is still uncertainty in costs and technical capabilities
of components. More details are available in [3]. For
the business-as-usual concept, offshore wind farms
are connected only through radial connections, in line
with the definition of the concept. On the contrary, in
the European decentralized wind power hubs concept,
synergies appear in the connection of Dutch and British
offshore wind farms, and of German and Danish offshore
wind farms.
Table 1 shows the costs for the two concepts, with the two
assumptions on the DCCBs, split per category. Note that,
due to the uncertainty in costs and technical capabilities
of components, these numerical results must be taken
with care. The European decentralized wind power hubs
concept allows important savings on submarine cables
compared to the business-as-usual concept. It is due to
potential synergies between the connection of different
offshore wind farms. It allows also small savings on the
 Figure 6. European decentralized wind power hubs concept: example of a possible resulting grid in 2030.

onshore converters. However, because the European
decentralized wind power hubs concept would require
DCCBs, it is very sensitive to the cost assumption for
these DCCBs: it could be either much less expensive (i.e.
by several billions) that the business-as-usual concept,
or much more expensive (i.e. by several billions). Table
2 gives then the additional investment of the European
decentralized wind power hubs concept compared to
the business-as-usual concept, but also the actualized
increase of the SEW over the lifetime (i.e. the generation
cost savings), and the Net Present Value obtained by
comparing the two numbers. Because the European
decentralized wind power hubs concept allows power
exchanges between countries through the offshore grid,
the annual variable generation cost is lower. However,
when expensive DCCBs are considered as necessary
(e.g. if technical capabilities of cheap DCCBs are not
sufficient to protect the system), it does not compensate
the additional investment cost.
5. Conclusions
With offshore wind energy, the North Seas can be viewed
as a potential reservoir of green electricity. Harnessing
these resources will be crucial to decarbonize the
European electricity sector and to reach the targets of
the European Union in terms of renewable energy.
Although several past studies show that the development
of an offshore grid is a cost-effective solution to unlock
the full potential of Europe’s offshore resources in the
North Seas, several technical, financial and regulatory
barriers are hampering the planning of such a grid.
The PROMOTioN project aims at alleviating these
remaining barriers. In that context, this paper contributes
to four subjects: (i) it proposes a set of planning criteria
that could be used for the development of a DC Meshed
Offshore Grid; (ii) it shows the results of a prefeasibility
study of a such a grid by looking at both the economic
and the technical feasibility of two specific concepts and
in particular, that the DCCBs capabilities and costs will
drastically impact the business case of meshed grids; (iii)
it shows the use of a new Optimal Transmission Expansion
Planning (OTEP) tool, allowing to derive automatically
grid topologies associated to different concepts; (iv) 3
different concepts are proposed that can be used for the
evaluation of technological and development roadmaps.
The PROMOTioN project is improving (e.g. costs
assumptions, grid protection strategies, topologies for

Table 1. Estimation of costs for the two concepts studied.

Table 2. Estimation of net benefits for the European decentralized wind power hubs concept.

Cigre Science & Engineering • N°13 February 2019

87
the various concepts) and enriching this prefeasibility
study to deliver a comprehensive Deployment Plan in
2019, such that a meshed HVDC offshore grid in the
North Seas can become a reality.
6. Acknowledgment
This work has received funding from the European
Union’s Horizon 2020 research and innovation program
under grant agreement No. 691714 (PROMOTioN).
7. Bibliography
[1] S. Cole, P. Martinot, S. Rapoport, G. Papaefthymiou, and V. Gori,
“Study on the benefits of a meshed offshore grid in the Northern
Cigre Science & Engineering • N°13 February 2019
88
Seas region,” European Commission, Tech. Rep., 2014.
[2] European Commission, “Commission Regulation (EU) 2016/1447
establishing a Network code on requirements for grid connection of
high voltage direct current systems and direct current- connected
power park modules,” 26 August 2016.
[3] O. Antoine, Y. Audichya, N. de Groot, P. Henneaux, K. Karoui,
M. Lessinnes, A. Mannocchi, S. Menze, O. Scheufeld, F. Velez
Terreros, and A. Wagner, “Deliverable 1.6: Draft roadmap
andreference offshore grid expansion plan,” PROMOTioN project,
Tech. Rep., 2017
[4] X.-P. Zhang, “Multiterminal voltage-sourced converter-based
HVDC models for power flow analysis,” IEEE Transactions on
Power Systems, vol. 19, no. 4, pp. 1877-1884, 2004. [5] C i g r é
JWG B4/C1.65, “Recommended voltages for HVDC grids,” 2017.
Icelandic operational experience of
synchrophasor-based fast frequency
response and islanding defence

D. WILSON1,5, B. HEIMISSON1, R. GUDMANSSON1, I. BALDURSDOTTIR1, O. ARMANNSSON2,

K. HALLDORSSON3, E. LINNET4, O. BAGLEYBTER5
1
Landsnet, 2Nordural, 3Landsvirkjun, 4ISAL, Iceland; 5GE Grid Solutions, UK

Summary relative values of time to respond and location sensitivity

The Icelandic power system experiences large and rapid
frequency deviations because of its low inertia and large
loads relative to the system size. Using fast synchrophasor
based signals to trigger new flexible demand- and
supply-side resources, fast response services can reduce
frequency deviations and the probability and impact of
islanding. This paper reports on the experience of the first
live operational fast response services addressing angle
and frequency stability. The approach improves reliability
and access to the grid in constrained locations.
There are flexible resources in supply and demand side
that can initiate power responses very quickly, and using
existing resources is more cost effective than deploying
new facilities such as storage. Incorporating them requires
flexibility and co-ordination by the Transmission System
Operator (TSO). As a partner in the MIGRATE European
project [1], Landsnet, the Icelandic TSO is field testing
fast-acting wide area control together with generation and
consumer-side collaborators.
Iceland’s electrical energy is supplied by 73% hydro and
27% geothermal resources [6]. Hydro generation can
be ramped, and geothermal resources have fast-valving
capability for severe disturbances. At the demand side,
large thyristor-controlled aluminium smelter loads can be
stepped up or down almost instantaneously, while other
loads can accept fast acting tripping. Fast synchrophasor
measurements are used for real-time identification of the
occurrence, impact and location of a disturbance, and
response is applied to counteract the stability impact. The
are discussed.
Other countries share concerns with low inertia. Inertia
is reducing in the GB system with increasing renewable
generation, and a similar flexible fast response capability
is being trialled [2], as well as a battery storage service [3].
South Australia has low inertia and a weak transmission
link to Victoria. A 100MW battery storage solution was
recently installed, providing fast frequency response
[4]. In Ireland, new ancillary services are defined to
reduce constraining-on synchronous generation [5].
The Icelandic approach using novel control of existing
capacity is relevant to many operators with concerns over
inertia or stability.
1. The need for fast response for
angle and frequency stability
Frequency stability involves containing the frequency
peak after a disturbance, and in Iceland this requires a
response to match the loss within about 4-7s. However, fast
response impacts angle stability, which defines whether
the system will split or remain synchronised. Accelerating
frequency response to act within the timeframe of the first
angle swing (typically around 1.5s in Iceland) requires a
control approach that is sensitive to the location of the
disturbance and the angle behaviour.
Figure 1 illustrates the conventional view of a frequency
disturbance and the action of primary and secondary
control to limit the frequency drop and restore near-
nominal frequency. Conventional primary frequency

* Douglas.wilson1@ge.com

KEYWORDS
Inertia, frequency control, synchrophasors, wide area control, stability, governor, demand response..

Cigre Science & Engineering • N°13 February 2019

89
 Figure 1 Conventional frequency response services
response is based on governor frequency control and is
typically delivered in 10-20s, limited by physical turbine
and governor constraints.
There are four concerns related to conventional control:
1. In common with other low inertia grids, Icelandic
frequency can reach very high values in a short time, as
shown by the 53Hz peak in Figure 3 around 4 seconds
after the disturbance. Turbine governors tend to be
too slow to arrest the frequency before the deviation
becomes very large.
2. Reducing the response time of conventional control will
tend to destabilise grid frequency; sustained oscillations
between 49 and 51Hz with 30 second cycle period have
been observed in Iceland and elsewhere [7].
3. Response within the first 2 seconds of a disturbance
interacts with the transient angle response and can
either improve or degrade the risk of islanding,
depending on where the disturbance occurs relative to
the response.
4. Response far from the source of the disturbances can
lead to increased loadflow and may trip overload
protection.
An example of a trip of a large load is shown in Figure
3. The acceleration is unequal across the grid, earlier
and faster close to the initial loss of load than at distant
locations in the grid.
This difference in frequency results in angle differences
Figure 2 Icelandic transmission system showing main 220kV centres of inertia
Cigre Science & Engineering • N°13 February 2019
90
increasing between east and west, and results in protection
tripping to avoid an out-of-step condition. In the event
in Figure 3, the network separates at around 1.2s from
the main disturbance, and frequency in the southwest
island reaches a peak of 52.8Hz four seconds after the
disturbance.
A fast frequency response of increasing load or reducing
generation is only helpful for the disturbance in Figure 3
if it is within the southwest area of the grid, i.e. in the area
of the disturbance. Applying response in the east would
increase the angle separation between the grid regions,
making overloads and splitting more likely, and producing
greater power imbalance in the resulting islands.
2. Wide area control mechanism
for fast response
The principle of using aggregated voltage phasors and
frequency to represent the coherent areas of the grid
is shown in Figure 4. The aggregated phasors are a
simplified and fault- tolerant way of representing the
power system in measurements that can be shared with
all fast response controllers. The controllers are activated
using these measurements, deriving a system-wide Rate of
Change of Frequency (ROCOF) for proportional response
(accounting for islanding status), using the aggregated
values to enable or inhibit response per area.
 Figure 3 Example of Load Loss Event leading to Islanding

The main elements of the control scheme are shown in Figure 5, with the following components:

Phasor Measurement Units (PMUs) Synchronised voltage and current measurements at 20ms updates. 45 PMUs are connected in
Iceland, of which 7 are currently used in the control scheme.

Regional Aggregator (RA) Aggregates the PMUs within each region in Figure 4, to produce a single angle and frequency
value representing the region. Aggregated signals are shared with all Resource Controllers. RAs
provide redundancy of measurements, reduced measurement noise, and better represent the region
movements better than individual PMU measurements.

Resource Controller (RC) Control units that receive the RA signals from around the system and process to provide triggering
signals to its connected resource(s). An RC is associated with a region, and there can be several
RCs in a region. It sends the trigger to the control interface at the plant through hardwired signals or
a fast protocol such as IEC 61850 GOOSE.

Control Interface to plant Plant control system that receives the trigger and translates it into actions.

Figure 4 Regional Aggregation of Frequency and Angles

Cigre Science & Engineering • N°13 February 2019

91
 Figure 5 Regional Aggregators (RAs) and Resource Controllers (RCs) in a distributed control scheme
Figure 5 shows the intended architecture where the RAs
share data with all RCs without going through a Phasor
Data Concentrator (PDC). However, for the trials, a simpler
initial architecture was used where data was sent from
the control centre PhasorPoint™ Wide Area Monitoring
System (WAMS) to units that act as RAs/RCs. This was
done to introduce wide area control in incremental steps
without impacting the WAMS infrastructure that is a
critical part of the operational system observability. The
performance in the trial is:
40ms PMU latency, consistent with P-class latency
in IEEE standard for synchrophasors [8]
160ms Network & PDC latency
>99.8% Data availability (control availability much
higher due to designed gap tolerance)
20ms Cycle time of PhasorController™ platform
20-300ms Algorithm latency, typically 20ms for digital
trigger, 80ms severe ROCOF, 300ms marginal
ROCOF event triggering
Further delay time and data availability improvements are
targeted through migration to the Figure 5 architecture,
principally by avoiding use of a monitoring PDC and by
enhancement of the event detection algorithm.
The process of determining the real-time triggering of the
system is as follows:
1. Data quality checks are applied to incoming data to the
RAs
2. Regional frequency and voltage phase angle signals
are produced by RAs using several PMUs. These
values are robust against loss or degradation of data
from some of the PMUs.
3 RA frequency and angle signals are shared with RCs.
Cigre Science & Engineering • N°13 February 2019
92
Each RC belongs to a region, and applies the following
process to the devices under its control:
a. Island detection. Identify if the system is islanded
using RA angle and frequency differences. The
RC determines which regions are connected to its
region.
b. System Aggregate values. The RC produces a
system frequency, system-ROCOF and system
angle across all of the connected regions.
c. Event trigger. The RC determines whether a high
or low frequency event has occurred, principally
by comparing system-ROCOF with thresholds,
which are adapted depending on which regions are
connected with the RC’s area.
d. Location enabling. Action is enabled or inhibited
for a region depending on whether the angle from
the RA in the relevant region is leading or lagging
the system average. Load reduction and generation
raising are enabled if the area angle is lagging and
frequency is low and/or falling. Load increase and
generation reduction are enabled in an area where
the angle is leading and frequency is high and/or
rising. Thus, angles are shifted towards the system
mean by the control action.
e. Direct trigger. Where appropriate, a very fast
initial response can be obtained and used in nearby
resources by triggering directly from a digital status
value, e.g. smelter potline operating trip. Locational
sensitivity must be respected.
f. Fast Ramp Duration. In the case of Generation
Fast Ramp, a triggering event will raise a signal to
indicate that the RC is controlling the machine and
a signal to command a generator ramp. The ramp
 signal stays high until the system need is satisfied,
e.g. downwards ramp until frequency reaches
50.2Hz. Further constraints are applied, such as
avoiding reverse power and rough running. Once the
ramp is complete, a further settling is applied before
the RC passes control back to the governor and the
standard governor operating mode is resumed.
4. Each RC reports its calculations and digital status
values into the WAMS system, and a real-time
operational summary is provided to the control room
operators, so that the operation can be supervised in
the control centre.
3. Characteristics of resources
The resources for fast control are categorized as discrete
or ramping control, as defined in Table 1. The resources
categorized as discrete control act rapidly to reduce
the slope of the frequency. This leads to the change in
gradient illustrated in Figure 6. Ramping responses use
the same trigger to start the ramp, and are subject to the
same locational enabling mechanism as discrete response,
but it takes time to deploy. The trigger signal is held on
as long as the ramping resource is needed. It can be
beneficial for the ramping responses to continue beyond
the zero frequency gradient since a hydro generator needs
to ramp through a rough-running zone, and because some
fast-acting resources e.g. smelters must return to their
original values. In the practical implementation, it was
decided to stop the ramp-down when frequency returned
within 50.2Hz.
In addition to the wide area controlled resources, there is
fast response delivered from geothermal power stations,
which can be very useful in containing frequency in
the south west. The geothermal fast-valving response
is short-duration, and turbines must return to their pre-
event output rapidly, and much of the response needs to
be replaced by hydro within about 6s. Another approach
being investigated (but not yet trialled) is to adapt the
line of splitting between network islands as they form, to
improve the power balance in the resulting islands. The
particular case investigated would choose which breaker
to open such that two hydro units would connect to the
East or the West islands.
In the trials, the resource profile for discrete response by
two smelters (Nordural and ISAL) is shown in Figure
6. There are two levels of upward power triggers (High
Frequency) and two downward (Low Frequency) for both
participating smelters. Figure 6 also shows measured
response where the resource was activated.

Examples Target Description

Latency
Discrete Smelter load step up or down 0.5s Response approximates a step in power, either by breaker
control (Nordural and ISAL) opening or power electronic control.
The smelter control is an examples of controlled load steps and
Load trip East Iceland Load Shed is breaker-tripped load shed. These are
(East Iceland Industrial Load) disctinct from conventional under/over-frequency tripping.

Ramp control Generator fast-ramp 2-20s Ramp control turns on within a target time of 0.5s and remains
(HRA hydro generator) on until a power balance is achieved (SROCOF=0) or continued to avoid
hydro rough running, or to pre-empt other
Load ramp short term fast response being restored. Down-ramping is
(future potential) trialled, but both directions could be used if available.

Thermal fast- Geothermal governor fast- 1-3s Currently triggered locally when f>51Hz, power can reduce in
valving valving about 2s to near-zero at all units. However, the turbine must restore power
quickly, typically returning to half-power within about 6s and continuing to
full power within about 1 min.
Intelligent Control decision between two <1.0 If the system response to a disturbance shows that the system
splitting breaker opening options to will separate into islands, the control scheme determines which connect 2
hydro units to east breaker to open so that 2x 55MW hydro units connect to the
or west islands. eastern or western islands.
Table 1 Categorisation of wide area responses

Cigre Science & Engineering • N°13 February 2019

93
 Figure 6 Smelter response profiles and measured responses at Nordural

Another discrete resource is load shedding in East Iceland.
Six fish processing factories with a total maximum load
capacity of 90MW are connected to a weak part of the
132kV network in East Iceland. The loads are tripped
for disturbances in which the east area is enabled by the
location function. Different triggering thresholds are
applied for each controlled resource, to so that the response
is proportionate to the event size. In one activation during
an islanding event, the load shedding scheme delivered
32MW of load shed response in the island, acting to
stabilise the frequency.
Ramping response is enabled at one 70MW unit at the
Hrauneyjarfoss (HRA) hydro station, with new governor
control functionality developed by Andritz. The interaction
between the wide area control and the governor control
is shown in a fast-ramp response to a real disturbance in
Figure 7. A full downward load ramp is achieved in just
17s, much larger and faster than conventional response.
The implementation of the HRA fast ramping approach
required design and commissioning of a new governor
software function that interacts with the wide area control
system. It also includes its own layer of operational
security to ensure that the unit operates within its
design capability for boundaries such as reverse power,
rough running (or “forbidden”) zone and ramp rate. It is
important that there is a valid power setpoint when the

Figure 7 Example of hydro governor fast ramp action in response to a grid disturbance

Cigre Science & Engineering • N°13 February 2019

94
control mode passes back to the normal operating mode.
The co-ordination between systems is designed and tested
to ensure that no situation can arise where the machine is
in an indeterminate control state.
As well as co-ordination between wide area control and
governing control, it is also important that the control
centre’s Automatic Generation Control (AGC) is co-
ordinated to ensure that the new operating point of the
generator is recognised without the potential of AGC
returning the unit to the pre-event power setpoint. A further
security measure to be added is a feature to recognise the
condition where wide area control data loss or equipment
failure results in the machine remaining in the rough
running zone, and control centre automation will move
the operating point away from rough running.
4. System response
The overall response profile of the non-conventional
response in south western Iceland is shown in Figure 8
for the large load loss shown within the area. The event
sequence is based on a real event, involving a large
load loss followed by a second smaller load loss and a
further load ramp, all within the southwest. The ideal
power balancing response would be immediate, close to
the volume tripped and sustained over time, so that there
Figure 8 Expected non-conventional power response contributions to western load loss balancing
Cigre Science & Engineering • N°13 February 2019
95
would be very little frequency and angle movement.
Clearly, this is not practically achievable, and there will
always be some frequency and angle movement due to
imbalance and need for some longer-term conventional
governing response. However, achieving performance as
close as practical to the ideal response reduces the impact
of the disturbance.
The profile of non-conventional response is shown
with the existing capability that is now active in the
southwestern part of the system. There is clearly a value
in rolling out the capability to further units to improve
frequency containment in the area, however the trials
already show the value of the control applied.
From Figure 8, it may be observed that:
• The thyristor-controlled smelter load ramp is the
fastest initial response in the southwest, addressing the
dynamic angle swing and reducing the frequency peak.
The control is implemented in both southwest smelters,
but the capability of one may be lost by a trip.
• The geothermal governor fast-valving response by
units in the southwest is very useful for containing the
frequency. There are potential improvements in starting
the response earlier for a close disturbance (before
51Hz is reached) with wide area control, and delayed
for a distant disturbance.
 Figure 9 Typical stages of response to a disturbance
• The geothermal response is short-term and must be
replaced.
• Hydro fast ramping response is sustained and replaces
the time-limited responses.
Applying this in more units would improve the overall
frequency behaviour.
• The load tripping resources are all in the East; the
action can be fast and sustained, but is also disruptive
for the customer and used only when essential.
5. Relative technical value of
frequency and stability services
In common with other TSOs addressing the issue of low
inertia and fast response, there is a need to encourage
flexible responses. The responses described in this paper
show different profiles, but the control approach enables
them to be co-ordinated to achieve an overall predictable
and effective response profile for each area. There are
considerable cost savings to be achieved in using existing
plants to manage the frequency and stability performance
rather than using dedicated resources such as battery
storage. To use existing resources, it is necessary to
account for inherent variability in responses arising
from the technical capabilities of different plants and the
operating state defined by the plants’ primary roles.
In the Icelandic situation, it is recognised that there is mutual
benefit to generation, demand and network stakeholders to
improve frequency and stability management, influencing
the security of supply and cost of ancillary services as
well as the stress and performance of plant. Also, there
is an environmental benefit in using the existing network
infrastructure more intensively, increasing the available
connection capacity while keeping network reinforce-
ment to a minimum. There is further work to be done in
the area of value and incentives in order to roll out more
Cigre Science & Engineering • N°13 February 2019
96
rapid response capability, but this is done in the context of
consensus across stakeholders that there is a shared value
in applying the approach more widely.
As a step towards defining flexible response services, it
is important to recognise the relative values of response
to the grid, accounting for their contribution to containing
frequency and preventing or mitigating islanding. Timing,
location, volume of energy delivered and ability to sustain
a response are important factors. Also, the frequency with
which a response may be called on may vary, for example,
load or generator tripping may be the fastest and most
sustained response but as it is the least desirable for the
primary function of the plant, it would be a ‘last resort’
reserved for the most severe disturbances.
Figure 9 illustrates the different stages of a disturbance.
Immediately following protection operation, there is a
dynamic angle swing, where angles will diverge typically
for about 1-2 seconds. The angles will swing back
together or the system will island. As illustrated in Figure
3, the frequency continues to rise until there is a balance
between generation and load and the frequency gradient
becomes zero. The frequency will return towards 50Hz,
but will not reach it by primary control alone. Secondary
mechanisms of Automatic Generation Control (AGC) and
operator dispatch will restore the system to its nominal
frequency.
Conventional services of governor-frequency regulation (ie
droop control) and spinning reserve address 4A, typically
starting to deploy at 3-4 seconds for hydro units and
continuing up to about 20-30s. However, the conventional
response has very limited capability to address stages 2
(Angle Stability) and 3 (Frequency Stability) because of
the shorter timeframes. Without fast response capability,
there is a risk of further loss of generation and load due to
islanding and local frequency protection of plant. Services
 Figure 10 Relative value of energy delivery in Stages 2-Angle Stability and 3-Frequency Stability

are needed to contribute to stages 2 and 3, but depending
on the technical characteristics, they may also be able to
contribute to stages 4A&B. Therefore, the duty for the
full range of stabilisation services can be shared across
generation and load.
Figure 10 illustrates the relative value of energy supplied
at different stages in a disturbance. The value of a 1MW.s
energy response delivery indicated by the green and
red injections is equal in terms of frequency stability as
long as it is delivered between the disturbance and the
frequency peak. Thus, the green and red injections of
energy response have the same effect on the frequency
peak although they occur at different times. Clearly, a
response that is intended to deliver a frequency stability
service can generally deliver a greater response if it
starts as early as possible. However, a frequency stability
response delivered in the wrong location during the angle
swing period has no value, or even negative value as it
increases islanding risk and may degrade the balance in
the islands if separation occurs. By contrast, the value of
1MW.s response to angle stability has a higher value if
delivered early, and the value diminishes up to the point of
the angle swing peak. There is only a value if the response
in the angle swing is location sensitive, otherwise it may
degrade the angle stability.
The technical value of the response therefore suggests a
time-dependent value profile as illustrated in the top chart
of Figure 11. There is additional value to location-sensitive
services that can respond within the first angle swing (ie
within about 3s) and also in services that respond before
the frequency peak (ie within about 4-7s).
Figure 11 shows the response profile of an example
participant, and the estimated value of the response.
Responses that can provide the early stage response
should be encouraged to do so, while those that cannot act
quickly can still derive value for a longer-term response.
Further work would be required if there is a need to
quantify the relative values of the time segments, and to
determine the volume of responses required for system
operation for each area.
Landsnet has a strong interest in probabilistic security
assessment, and demonstrated a novel on-line security
assessment toolset in the EU GARPUR project [9,10].

Figure 11 Illustrative response value profile for stages of disturbance response

Cigre Science & Engineering • N°13 February 2019

97
Using the methodology, it is possible to assess the
probability and impact of outages with and without
smart grid technologies. While the approach currently
has an emphasis on steady state loading, there is an
interest in the approach to dynamic issues, with the aim
of quantifying the benefits of risk reduction due to fast
response capability.
6. Conclusion
The technical demonstration proves that a fast,
locational response to grid disturbances is possible,
and that it can be achieved with a distributed control
approach using synchrophasor measurements. The
control approach is designed to utilize various different
response technologies, both load and generation, to
provide a coordinated response without adding new
power equipment.
The benefits include:
1. Reducing the occurrence of islanding and the
consequent risks and losses of generation and load.
2. If islanding does occur, it improves the likelihood
of the islands surviving, as there is an improved
balance in the islands as they form.
3. Reducing the risk of load or generation loss through
extreme frequency deviation.
4. Increased efficiency in delivering frequency
response services by opening the service to many
more participants, without incurring the cost of
installing dedicated devices.
5. Resources can be used that connect at any voltage
level, including transmission and distribution levels.
6. Reduced stress on machines that are involved in
frequency control:
a. Reducing mechanical stress from large speed
excursions during disturbances
b. Reducing the time of operation within the rough
running zone
c. Reducing continuous frequency control action;
separating large-disturbance response from
small-disturbance provides scope for reducing
oscillations.
7. Increased transfer capability of the network, as a
result of actions that reduce angle separations.
Iceland’s experience of operating a system with low
inertia is valuable for other systems where there is
low or reducing inertia. It is also relevant for larger
Cigre Science & Engineering • N°13 February 2019
98
interconnections where parts of the system have low
inertia and where there are transmission boundary
constraints. The novel technical approach leads to the
opportunity to contribute services that combine angle
stability, frequency stability and balancing, provided
by any grid users who have a useful response capability
to offer.
It is envisaged that more resources will be added to
the wide area control system using the mechanisms
described as templates for new additions to the scheme.
An extension to include geothermal fast-valving within
the wide area control scheme would be valuable at the
geothermal plants. Future work is also planned on the
networking architecture to further improve latency
and availability of the wide area control system as the
operational dependence on the system increases. It
is also intended that the work on intelligent network
splitting should be generalized and applied to other
areas of the grid where there is a risk of islanding,
particularly as new generation is added to the weak
network in the north.
7. Acknowledgements
The authors gratefully acknowledge the contributions
of Andritz collaborators, Florian Manz and Thomas
Haag, in developing and installing the hydro governor
capability and the personnel at Landsvirkjun, in
particular Kristján Björgvin Bragason and Axel
Sigurðsson, on the commissioning tests and generator
trials.
The funding support from the European Commission
under the Horizon 2020 MIGRATE project is
acknowledged.
8. Bibliography
[1] European Commission, Horizon 2020 project MIGRATE https://
www.h2020-migrate.eu/ accessed 26/1/2018
[2] Wilson D. H., et al: “Advances in Wide Area Monitoring and
Control to address Emerging Requirements related to Inertia,
Stability and Power Transfer in the GB Power System”, C2-208,
CIGRE Session, Paris 2016
[3] National Grid, “Enhanced Frequency Response (EFR)”, https://
www.nationalgrid.com/uk/electricity/balancing-services/
frequency-response-services/enhanced- frequency-response-efr
retrieved 25/1/2018
[4] Climate Home News, http://www.climatechangenews.
com/2018/01/11/teslas-south-australian-super-battery-beats-
 expectations-first-month/ 11/1/2018, retrieved 25/1/18
[5] Eirgrid: “DS3 System Services: Portfolio Capability Analysis”,
2014, http://www.eirgrid.ie/site- files/library/EirGrid/DS3-
System-Services-Portfolio-Capability-Analysis.pdf
[6] Energy Statistics in Iceland, 2016, http://www.orkustofnun.is/
gogn/os-onnur-rit/Orkutolur-2016-enska.pdf retrieved 29/1/2018
[7] O. J. Arango, et al: “Low Frequency Oscillations in the Colombian
Power System – Identification and Remedial Actions”, CIGRE
Session, Paris 2010
Cigre Science & Engineering • N°13 February 2019
99
[8] IEEE C37.118-2011(rev 2014): “Standard for Synchrophasors for
Power Systems”, 2014
[9] European Commission, Horizon 2020 project GARPUR, www.
garpur-project.eu accessed 26/1/18 [10] Perkins S., “Real-time
Weather-dependent Probabilistic Reliability Assessment of the
Icelandic Power System”, PhD thesis, Reykjavik University, 2018
[10] Perkins S., “Real-time Weather-dependent Probabilistic Reliability
Assessment of the Icelandic Power System”, PhD thesis, Reykjavik
University, 2018
Use of avian radar as a method to study
the effect of bird flight diverters
F. B. JOHANSEN1, S. S. INDERBERG, L. V. HAMMER
Norwegian Water Resources and Energy Directorate, Norway
Summary
High voltage transmission lines may cause fatal injuries
to birds, due to either electrocution or the bird colliding
with the lines. In this project, we wanted to study the
effect of bird flight diverters on collision rates. We
wanted to test if it was possible to use a bird radar to
register birds’ flight and behaviour, and this way get data
on the effects of bird flight diverters.
The project gave useful experiences about the use of
radars to collect data about birds’ response to bird
flight diverters. The analyses of the data indicated a
positive effect of using bird flight diverters as a method
to decrease fatal bird collisions. The results showed a
significant lower number of flight tracks in short distance
to marked lines compared with unmarked lines. In
addition, the turning angle of the birds were significant
higher near marked lines compared with unmarked lines.
This indicates that use of bird flight diverters on the top
line increase the visibility of the lines for birds, and as a
response to this, birds adjust the flight direction to avoid
impacts.
This project also have given useful experiences about
the use of avian radar as a method to investigate the
effects of bird flight diverters on power lines. The radar
registered the birds’ flight in an accurate way, and this
shows that the radar could be used in an extended study
on the topic in the future.
1. Introduction
A large number of birds dies every year due to collision
with power lines. In the US it is estimated that between
8 and 57 million individuals are killed annually [1]. The
number for Canada is believed to be between 2,5 and 25,6
* fbj@nve.no
KEYWORDS
Powerline - Birds - Collision - Marking - Method - Bird Flight Diverters - Spirals - Avian Radar
Cigre Science & Engineering • N°13 February 2019
100
million individuals [2]. In Norway, a study carried out
by the Norwegian Institute for Nature Research (NINA)
along a 300 kV power line in Ogndalen, Nord-Trøndelag
(2011-2013), showed that collisions with the unmarked
power line reduced the population of capercaillie
(Tetrao urogallus) with 4,2-11,8% and the population
of black grouse (Lyrurus tetrix) with 1,4-8,1% . The
study also showed that small birds down to common
redpoll (Acanthis flammea) collides with power lines
[3] Marking of power lines with bird flight diverters to
reduce the extent of bird collisions have therefore been
an important topic for many years in Norway.
Some research projects in South Africa, Europe and the
US have showed that physical enlargement of the phase
and/or earth wire due to some sort of markings could
result in a significant reduction in the number of birds
colliding with the power line. In the US, the frequencies
of collisions of cranes and ducks were significantly
reduced in sections where the power line was marked
[4]. Birds of prey also seems to react on marked power
lines with an increase of flight altitude to avoid collision
with the power line [5]. A study from Austria and
Hungary showed that the number of great bustard (Otis
tarda) colliding with a power line were significantly
reduced when the power line was marked [6]. No study
investigating the direct effects of marking of power
lines to reduce bird collisions has been done in Norway.
One of the reasons for this is because it is difficult to
measure exactly how birds react to the diverters. If a
bird collide with the line, it will likely fall dead to the
ground somewhere in the vicinity of the power line, and
the carcass will often quickly be eaten by scavengers.
This makes it difficult to estimate the extent of fatal bird
collision near power lines in an ecosystem dominated by
forest and with carnivores and scavengers present. It also
makes it difficult to test whether bird flight diverters are
an effective tool to reduce the number of birds colliding
with power lines.
Birds colliding with power lines could cause outages,
which are undesirable due to demand of security of
supply and economical consequences. In addition,
higher mortality could have biological and species
specific consequences. Bird collisions have proven to
be dependent of species, location and season, but due
to relative high costs of marking and uncertainty effect
of this, there are still many questions around this issue.
When the Norwegian Water Resources and Energy
Directorate (NVE) give a license to build and operate
a power line, one of the most important considerations
around biodiversity is which effect the power line will
have on birds. Since we have no conclusions about the
effects of marking of power lines in Norway, NVE often
decides that the power line should be marked with bird
flight diverters in bird rich areas – as a precaution to
decrease the number of birds that are killed in collisions
with the wires. The areas where the power lines are
marked could be in areas where the power line crosses
wetlands, marches, rivers, lakes, valleys and other bird
hot spots. However, use of bird flight diverters costs
money (approx. 2 - 3000 euros per kilometre if the
markers are mounted during the construction phase),
and makes the line more visible to humans. There
are also technical issues to consider when using bird
flight diverters. The diverters can cause problems with
corrosion due to salt building up under them in coastal
areas, and they accumulate ice, which can be a severe
problem in a cold environment such as in Norway. On
the other hand, marking can possibly solve problems
with birds being killed in collisions with the power lines,
and by reducing the number of collision, the number of
outages can also decrease. Therefore, it is important to
know whether the markings are effective or not, since
general marking costs money and have technical issues.
Until now, NVE have assumed that this kind of general
marking of power lines could reduce the number of birds
killed in collisions.
However, since no research has been done on this topic
in Norway, more information about this was desired.
Cigre Science & Engineering • N°13 February 2019
101
Therefore, NVE contacted the Norwegian Institute for
Nature Research (NINA) to do a study on the effect of
marking power lines with bird flight diverters.
In this pilot study we wanted to test whether we could
use an avian radar to reveal if marking the power lines
could affect 1) bird flight behaviour and 2) the frequency
of birds passing near the line. Our hypothesis was that
marking a power line with bird flight diverters could
1) make birds avoid marked sections of line more than
unmarked sections, and that 2) the frequency of birds
passing near the marked section would be lower than the
unmarked section. The main focus of the pilot study was
to find out if an avian radar could be a suitable tool to
monitor the bird activity around power lines, and hence
investigate the effect of bird flight diverters.
Parts of this paper is based on the final report for the
pilot study [7]. All figures are copied and translated from
the same report.
2. Method
2.1 Selection of area
It is well known that the number of birds colliding
with power lines will vary with topographic structures
in the landscape. Topography is also important for the
radars functionality. For an optimal use of avian radar in
projects like this, it is important to choose areas that are
relatively flat. In addition, the area should be near a road,
so that the car-mounted radar could have clear sight to
the power line.
After having consulted with the Norwegian TSO,
Statnett, a section of the Ormen Lange-power line was
chosen. The registrations was conducted during the
autumn migration in 2016 (August 17th- October 31st) in
an area near Molde in Møre og Romsdal on the west
coast of Norway (figure 1). Here a 420 kV power line
crosses the Os river, and the earth wire of the section
which crosses the river are marked with grey spirals
approx. 10 meters apart (figure 2 and 3). The reason
why the power line is marked on this section is a high
number of duck, cranes, swans, herons, geese and white-
tailed eagles flying along the river [8]. The position of
the radar made it possible to monitor both the marked
and the unmarked section at same time.
 Figure 1. Map showing the position of the avian radar (red dot) and the unmarked section (blue line) and
marked section (red line) of the power line. The power lines poles are marked with green dots.

2.2 The avian radar
NINA have used the avian radar in different projects
since 2008. The radar is a ROBIN 3D Flex, and is
mounted on a truck to make it movable (figure 4). This
mobile setup is designed by NINA to monitor birds in
three dimensions over a longer period at time. The radar
can monitor birds’ presence and behaviour over large
areas with high accuracy day and night throughout the
year. The radar is developed by Robin Radar Systems in
the Netherlands, and consists of a X-band based FMCW-
radar (Frequency Modulated Continuous Wave) and a
S-band radar designed for automatic surveillance and
tracking of birds. The S-band radar is a standard marine
navigation radar from Furuno, model FAR2167DS. The
FMCW-radar is mounted on a 2.5 meter high pole when
it is running, and rotates both horizontally and vertically.
It is used to measure the height of birds’ flight, and have
a maximum range of 3.5 kilometres. The FMCW-radar
is a “solid state coherent radar” with a dual sender/
receiver- antenna, specially designed and built by Robin
Radar Systems. It can track the altitude of birds’ flight
in all directions.
To make sure that a proper spot was chosen for the radar,
a drone was flown around in the area to test the radars
ability to track it. The test showed that the radar was
able to track the drone properly, but revealed some spots
where the drone was hidden for the radar due to high
trees shading for the radar waves. After this test, the
radar was moved to another spot where the radar had a
clearer view towards the power line.

Figure 2. One of the two earth wires are marked with spirals Figure 3: The earth wire was marked with grey plastic spirals
(photo: Bård G. Stokke, NINA). The picture shows the distance (photo: Roel May, NINA).
between the markers.

Cigre Science & Engineering • N°13 February 2019

102
 Figure 4: The radar mounted on a truck, with the S-band radar on the roof (left photo) and the FMCW-radar placed on the ground
(right photo) (both photos: Bård G. Stokke, NINA).
2.3 Visual verification
The avian radar registered bird flight routes and
frequency of birds passing the line, but since the birds’
behaviour could be species specific, it was also important
to conduct visual registrations in addition to the radar
monitoring.
Visual verification of radar tracks from birds was
conducted September 12th and October 3rd, 4th and 13th.
Visual observations of birds were then compared with
tracks from the avian radar. Species and number of birds
were recorded in the system. This verification showed
that the radar registered the birds’ flight in a good way,
but unfortunately, data of flight height (from the FMCW-
radar) lacked for some of the recorded tracks.
2.4 Database and data processing
The database with registered tracks was copied once a
month and saved in a PostgreSQL database program.
The processing of the data consisted of sorting out tracks
that was not from birds and classifying birds in groups
(flocks, large, medium sized and small birds). The data
was analysed in birds’ presence in time and space,
frequency of birds crossing the power line and flight
behaviour in vicinity of the power line. Flight behaviour
included an analysis of changes in turning angle as a
function of distance to marked and unmarked section
of the power line. We expected the turning angle to be
larger near marked sections compared with unmarked
section of the power line.
Potential effects of the markings with bird flight diverters
was done by:
1) Differences between day and night: Analysing tracks
on marked and unmarked sections compared between
day and night. Due to lack of daylight at night, we
expected no effects of the marking at night time and an
effect at daytime.
Cigre Science & Engineering • N°13 February 2019
103
2) Distance to the wire: Comparing tracks near marked
and unmarked sections by measured distance to the wire.
If the bird could see the markings, we expected that it
should be fewer tracks near the marked section.
3) Turning angle: Analysing the birds’ turning angle near
marked and unmarked sections. We expected a bird that
fly near a marked section to show evasive action when
seeing the marking.
2.5 Statistics
Potential effects of the markings was analysed by using
a BACI- and a gradient approach. In the BACI approach,
tracks by unmarked and marked section were compared
at daytime and nighttime. In the gradient approach,
tracks by marked and unmarked sections were compared
by measured distance to the wire. For all approaches,
four models were compared using information theoretic
approach [9].
Flight direction was cosinus transformed before the data
was analysed to achieve a normal distribution, and was
defined as a response variable in a general mixed effects
models with date as a random grouping factor.
The probability of bird tracks crossing the power line was
calculated as the number of flight tracks in a crosswise
direction with the power line (± 45°) compared with
the number of tracks in a lengthwise direction with the
power line. This was modelled using a generalized mixed
effects model with binomial distribution and “date” as
random grouping factor.
The number of radar tracks by birds per day was analysed
using a general mixed effects model with Poisson
distribution and “date” as a random grouping factor.
The turning angle of the bird tracks was calculated as
(1-cos(∆direction))/2. This gives values form 0 (no
change in turning angle (0°)) to 1 (full U-turn (180°)).
Turning angle was used as a response variable in a linear
mixed effects model with “date” as a random grouping
factor.
The modelling was done using lmer and glmer in the
lme4-library [10] in R 3.3.2 [11].

3. Results
3.1 General
In general, the verification process showed that the
radar registered bird tracks in an accurate way, however,
some of the data that showed the birds’ flight height
lacked. It will have to be checked with the radar supplier
why this happened. In addition, some tracks for birds
that fly low (below 50 metre) was not registered. One
possible explanation is tall trees shading for the radar. Figure 5: The figure shows the number of tracks during day and night for the
The verification also showed that both large flocks marked and the unmarked section.
of birds (like fieldfare (Turdus pilaris) and redwing
(Turdus iliacus)) and larger birds (like mallard (Anas
platyrhynchos), great cormorant (Phalacrocorax carbo) 3.3 Effects of distance to the power line
and grey heron (Ardea cinerea)) passed the power line
The number of tracks near the marked section decreased
and was registered in the database.
more with lower distance to the wire for the marked
After the registration period was over, the data was section (F=1289.239) than for the unmarked section
cleaned to remove tracks from other objects than birds (F=10052.733, figure 6). On a 50-metre distance to the
(like cars and pedestrians). The total number of tracks wire, it were significantly fewer tracks by the marked
from birds in a 200 meter distance from the power section than by the unmarked section. This probably
line was 40 722 at the unmarked section, and 37 784 means that the birds are aware of the markings, and keep
at the marked section. The activity decreased after mid- a sufficient distance to the power line.
October, probably because the autumn migration was
near the end. Within a distance of 200 meters from the
power line there was little difference in the radar’s ability
to detect objects near marked and unmarked sections,
and this shows that the data for the two sections are
comparable and suitable for further analysis.
3.2 Effect of time of day
The number of tracks were significantly lower in
daytime compared with nighttime near the unmarked
sections (F=2415.87), but such significant difference
was not found for the marked sections (F=711.69). If
the marking with bird flight diverters had an effect, it
was expected that more birds passed the marked section
compared with the control section during nighttime
(because the markings are only visible at daytime), but
this did not happen (figure 5).
Figure 6: There are fewer numbers of tracks near the marked section of the
power line compared with the unmarked section.

Cigre Science & Engineering • N°13 February 2019

104
3.4 Frequency of birds crossing the power line section (F=958.09), especially at daytime (F=24.901).
This indicates strongly that the birds observes the bird
The probability that a track crossed the power line was
flight diverters and adjust their flight behaviour near the
significant lower by the marked section compared with
marked section of the power line (figure 8).
the unmarked section (F=447.120). Also, if one look
at the distance from the power line, the probability
of tracks crossing the power line was higher by the
unmarked than the marked section in a large distance
from the line, but approximately the same at a distance
of 50 metres (F=349.993, figure 7). These results show
that the birds crosses the marked section in lower rate
than the unmarked section. This supports our hypothesis
that the birds see the bird flight diverters and adjust
their flight behaviour if the see the markings in a long
distance. However, if the bird flies near the power line
(50 meters or less), they don’t show the same adaptive
response toward the markings.
Since the part of the radar that monitor the birds’ height
didn’t work as expected, many tracks lacked information
on altitude. This makes it impossible to conclude about
how the bird alter their flight behaviour near the power
line. One possibility is that when the birds cross the marked
section they alter their altitude and cross the power line in Figure 8: The figure shows birds’ turning angle near marked and unmarked
section of the power line.
a safe altitude, but at this point it is just speculations.
4. Conclusion
This pilot project have given useful experiences about
the use of avian radar as a method to investigate the
effects of bird flight diverters on power lines. The radar
registered the birds’ flight in an accurate way, and this
shows that the radar could be used in an extended study
on the topic. However, we will have to address the
difficulties with the part of the radar that registered the
birds’ altitude before further studies are done.
The analysis of the data indicates that there is a positive
effect of using bird flight diverters on the earth wire
to reduce the risk of birds colliding with the wire. A
significantly reduction of bird tracks near the marked
section of the power line indicates that the birds are
aware of the markings. The most interesting result was
that the birds’ turning angle was significantly higher
Figure 7: The figure shows the probability that the tracks crosses the marked by the marked section compared with the unmarked
and unmarked section in four different distances to the power line.
section. This is a clear indication that the use of bird
Turning angle flight diverters on the earth wire increase the visibility
of the power line for birds.
The turning angle of the birds was significantly higher
by the marked section compared with the unmarked This pilot study has been carried out in an area where

Cigre Science & Engineering • N°13 February 2019

105
power lines have been established for a long time. The
birds’ behaviour and activity might have adjusted to these
structures over time. We will start a new project based
on this pilot study to get clearer answers about how birds
adjust their flight behaviour near power lines that are
marked with bird flight diverters. This study will have
a BACI- approach (before-after, control-impact) where
the radar will be used to monitor areas where new power
lines are built, with marked and unmarked sections. The
new project is a cooperation between NINA, Statnett and
NVE, and the results from this study will hopefully give
a clearer understanding of how bird flight diverters can
be used to decrease bird colliding with power lines. The
study will be carried out between 2018 and 2021, and is
financed by the Research Council of Norway, Statnett
and NVE. The total budget is 1,1 million euros.
5. Bibliography
[1] Loss, S.R., Will, T. & Marra, P.P. 2014. Refining estimates of bird
collisions and electrocution mortality at power lines in the United
States. PLos ONE 9: e101565.
[2] Rioux, S., Savard, J.-P.L. & Gerick, A.A. 2013. Avian mortalities
due to transmission line collisions: a review of current estimates
and field methods with an emphasis on applications to the
Canadian electric network. Avian Conservation and Ecology 8: 7.
[3] Bevanger, K., Bartzke, G., Brøseth, H., Dahl, E.L., Gjershaug,
J.O., Hanssen, F., Jacobsen, K.- O., Kleven, O., Kvaløy, P., May,
Cigre Science & Engineering • N°13 February 2019
106
R., Meås, R., Nygård, T., Refsnæs, S., Stokke, S. & Thomassen,
J. 2014. Optimal design and routing of power lines; ecological,
technical and economic perspectives (OPTIPOL). Final Report;
findings 2009 – 2014. NINA Report 1014. 92 pp.
[4] Brown, W.M. & Drewien, R.C. 1995. Evaluation of two power
line markers to reduce crane and waterfowl collision mortality.
Wildlife Society Bulletin 23: 217-227.
[5] Luzenski, J., Rocca, C.E., Harness, R.E., Cummings, J.L., Austin,
D.D., Landon, M.A. & Dwyer, J.F. 2016. Collision avoidance by
migrating raptors encountering a new electric power transmission
line. Condor 118: 402-410.
[6] Raab, R., Schütz, C., Spakovszky, P., Julius, E. & Schulze, C.
H. 2012. Underground cabling and marking of power lines:
conservation measures rapidly reduced mortality of West-
Pannonian great bustards Otis tarda. Bird Conservation
International 22: 299-306.
[7] Stokke, B. G., Åstrøm, J., Bevanger, K., Hamre, Ø. & May, R.
2017. En radarbasert undersøkelse av effekten på fugl av merking
av høyspentledninger. NINA kortrapport 59. 23 pp.
[8] Stenberg, I. 2010. Nettilknytning Ormen Lange. Biologiske
etterundersøkingar. Årsrapport 2009. 69 pp.
[9] Burnham, K.P. & Anderson, D.R. 2002. Model selection and
multimodel inference: a practical information-theoretic approach,
2nd edn. Springer, New York, NY, USA.
[10] Bates, D., Maechler, M., Bolker, B. & Walker, S. 2015. Fitting
linear mixed-effects models using lme4. Journal of Statistical
Software 67: 1-48. doi:10.18637/jss.v067.i01.
[11] R Core Team. 2016. R: A language and environment for statistical
computing. R Foundation for Statistical Computing, Vienna,
Austria. (https://www.r-project.org/).
Measurement and analysis of harmonic
data to assess the impact of installations
connected to high voltage systems

F.T.J. VAN ERP1, .F.G. COBBEN2, R.A.C.T. DE GROOT1, J.B.M. VAN WAES1
1
TenneT TSO B.V., 2Eindhoven University of Technology
The Netherlands1,2

Summary harmonic disturbance levels in power supply systems.

Inverter-based technologies are increasingly applied in
high and extra high voltage systems. In addition to that,
overhead lines are replaced by underground cables due
to spatial planning needs of society. The use of more
underground cables causes changes in the resonance
frequencies of the power supply system impedance, which
can result in either amplifying or reduction of harmonic
levels in power supply systems.
A practical harmonic assessment of the impact of a
distorting installation on an extra high voltage grid is for a
real installation complicated. Articles are published on the
methods, but not many describe real cases in high voltage
grids. This paper analyses a harmonic case with actual
measurements. The analysis is based on the non-invasive
method [1]. In this particular case, when the disturbance
originates from the grid, an assessment can be performed
with known filter states and calculated filter impedances.
The 7th harmonic has its origin in the grid. In case of the
other harmonic orders it cannot be concluded that, either
the disturbance is due to the grid, or due to the distorting
installation, or both. In these cases the power directions
varies in time.
1. Introduction
Inverter-based technologies are increasingly applied
by customers connected to high and extra high voltage
systems. In addition to that, overhead lines are replaced
by underground cables due to spatial planning needs
of society. The use of more underground cables causes
changes in the resonance frequencies of the power supply
system impedance, which can result in higher or lower
The IEC document [2] specifies guidelines for emission
limits for customers.
To control the harmonic disturbance levels in power
supply systems, organizations responsible for planning
and operating the power supply systems provides limits
for planning purposes and emission limits for use during
operation. A customer connected to the power supply
system calculates emission levels for its distorting
installation, which are assessed by organizations
responsible for the planning and operation of the power
supply system.
In addition to reviews of the theoretical determined
emission levels, measurement methods and analysis
methods are needed to assess the emission levels. These
methods shall provide insight whether a disturbance
originates from the power supply system, customer or
both. This paper is intended to share experience with
a practical case study with actual measurements and
analysis. It is based on a noninvasive method.
This article starts with a brief description of the used
method. The measurement setup is explained in the chapter
‘Measurement distorting installation’. It is followed with
the main measurement results and analysis in the chapter
‘Measurement results and analysis’. Finally this article is
completed with the chapter ‘Conclusions’.
This paper is using the power directions of the harmonics
as basis for the method. This direction is related to the
interaction between the grid and the installation at the point
of evaluation. The interactions such as resonances and
cross-amplification of harmonics will not be considered
in this paper.

* Frans.van.Erp@TenneT.eu

KEYWORDS
Power quality, harmonic assessment, distorting installations, harmonics, power, phasor.

Cigre Science & Engineering • N°13 February 2019

107
2. The method used
This section describes the method in [1]. Figure 1 and
Figure 2 give a schematic representation of the grid (g)
and the distorting installation (c) of the customer.
Ūhg : the harmonic voltage phasor of the grid;
Ūhc : the harmonic voltage phasor of the distorting
installation.
The Thévenin circuit diagram has the relationships as
expressed by formula (1) per symmetrical component.

(1)

Figure 1: Thévenin circuit diagram The Norton circuit diagram has the relationships as
expressed by formula (2) per symmetrical component.

(2)

Figure 2: Norton circuit diagram

With:
Zhg : harmonic complex impedance of the grid; Figure 3 shows three typical cases of measured harmonic
Zhc : harmonic complex impedance of the distorting voltages versus the current. The areas reflect the
installation; measurement dots.
Ūh : the harmonic voltage phasor at POE;
Īh : the harmonic current phasor at the POE; Three cases can be distinguished:
Sh : apparent power phasor at the POE; • The blue area between the slopes of Figure 3a
Ph : harmonic active power of Sh; describes a situation that the harmonic current and the
Qh : harmonic reactive power of Sh; resulting voltage are due to the combined influence
Īhg : the harmonic current phasor of the grid; of the background level and the distorting installation
Īhc : the harmonic current phasor of the distorting without any prevalence of a one or the other.
installation; • In Figure 3b the data points are grouped around the

Figure 3: Typical plot area of meas ured harmonic voltage vs . harmonic current at the POE

Cigre Science & Engineering • N°13 February 2019

108
 impedance of the distorting installation. The grid is the
dominant harmonic source.
• When the data point are grouped around the slope of
the grid impedance then the distorting installation is
the dominant harmonic source, see Figure 3c.
Long duration simultaneous measurements of harmonic
voltage and current are used for the analysis. The
correlation between the harmonic current and voltage at
the point of common coupling is studied. The relations
between the measured values, the parameters of the
model, the changes of the known model parameters
and the (measurement) uncertainties are included. The
analysis is based on steady state conditions.
3. Measurements distorting
installation
The considered distorting installation is a 380kV HVDC
connection in The Netherlands. It includes a filter bank.
Depending on the operational conditions of the HVDC
installation, the filter modules are connected in parallel
to the grid. There are five filter states of which four are
dominant. Each state resembles a harmonic filter impedance
of a specific order. There is a bi-directional power flow.
The analysis is based a non-invasive method, i.e. study
is based on the natural response of the installation.
Simultaneous measurement of harmonic voltage and
current with phasor information are carried out at the
point of connection (POC). A power quality measurement
system (PQMS) measures the Ūh and Īh at the primary
terminals of the POC. These primary current and voltage
phasors are measured over a ten minute time interval.
The PQMS consists of resistive-capacitive voltage
transformers (RCVT) [3], current transformers (CT) [4],
and a measurement unit (MU) [5].
For a correct assessment of the results insight in the
accuracy of the measurement chain is required. The
RCVT has a voltage (ratio) error (εu) less than 0,2 %
and a phase displacement (Δφ) of less than 1° up to and
including 20 kHz. The CT has a current (ratio) error of
class 0,2S and the MU has a class
A (Maximum error = ±5% Um, details can be found in
[5]). Based on these measurement devices the combined
relative rectangular distributed uncertainty [6] of the
harmonic voltage (Uh) is estimated in Figure 4.
Cigre Science & Engineering • N°13 February 2019
109
Figure 4 relative uncertainty of the meas ured harmonic voltage Uh (|Ūh|)
Uh : measured voltage reading at POC;
Upr : rated primary voltage [7] .
In this typical case one percent measured reading
has approximately 5% combined relative rectangular
distributed uncertainty (εuh) of the harmonic voltage (Uh)
for the complete measurement chain.
To correlate the harmonic line voltages and current
phasors, both harmonic phasors are symmetrically
transformed to harmonic positive and negative currents,
voltages, and power phasors.
4. Measurements results and
analysis
This chapter shows in what way the measurements are
selected that have no impedance state changes of the
distorting installation. Next are shown the measurement
magnitudes of the harmonic voltages in a histogram to
evaluate the measured magnitude against the measurement
uncertainty. This chapter also shows the operational
conditions of the installation by means of symmetrical
fundamental power phasors. The measurements were
taken over period of approximately seven months to
include many operational states of the grid and the HVDC
installation. A selection of harmonics is used to illustrate
the results.
Distorting installation impedance state changes
The filter modules influence the impedance of the
installation (Zhc). When the filter state changes during
the ten-minute measuring interval, the harmonic voltage
(Ūh) and impedance changes (ΔZhc). The intervals with
 Figure 6: magnitude harmonic s ymmetrical voltage |Ūh|
changes between two steady state conditions are excluded
from the analysis.
This results in ten-minute measurements with more
constant filter impedance (Zhc) per filter state. In Figure 5
an impression is given of the calculated filter impedance
and the moments that the multi module filter is turned
on and off. The green arrows in Figure 5 and green dots
in Figure 9 indicate the measurements that are excluded
from the measurements. Based on the design values the
filter impedances where calculated.
Figure 5: The green arrows indicate the interval in which the s tate
of the filter changes
Measurement accuracy
In Figure 6 the voltages at the POC are depicted. The
PQMS measures line-to-line harmonic voltages and phase
currents. To correlate the harmonic voltages and current,
both are transformed to positive (ps) and negative (ns)
symmetrical components.
Cigre Science & Engineering • N°13 February 2019
110
The figures show that the symmetrical voltages are slight
above, and below 1% of Upr. When the voltage Uh is less
than 1% of Upr, the relative uncertainty increases, see
Figure 6 dark area.
Apparent power
In Figure 7 the symmetrical power phasors at the
fundamental frequency are presented for various operating
conditions. The rated power of the distorting installation
is one per unit.
The negative active power means that the installation is
producing active power. The positive reactive power is
absorption of reactive power.
The symmetrical harmonic power phasors are depicted in
Figure 8. It shows that;
• the phase angle of the 5th and the 7th harmonic are
independent of operating conditions;
• the direction of the power flow of the 13th harmonic
varies in time.
Figure 8 shows dominate negative sequence harmonic
power at harmonic orders (6k-1), and dominant positive
sequence harmonic power at harmonic orders (6k+1), as
6 pulse current source investors would produce.
Currents and voltages
In Figure 9 the long duration simultaneous measurements
of harmonic voltage and current [1] are depicted in
symmetrical components.
Figure 9 shows that:
• the 7th harmonic there is a linear relation between |Ūh|,
and |Īh|, as well as Ph, and Qh of Figure 8;
• the 5th, and 19th there is little correlation between the
|Ūh| and |Īh|;
 Figure 7: operational condition cus tomer: 50 Hz s ymmetric apparent power phas ors

• the 13th has a more or less reversed correlation between
the |Ūh| and |Īh|.
Based on Figure 3 [1] the origin of the harmonic source
can be identified. Analysing the measurement results the
findings are:
1. Based on figure 8 the grid is the dominant harmonic
origin for the 7th harmonic: in Figure 9, the points of
the seventh harmonic order lie along with the four
straight lines with a slope approximately equal to
modulus of the filter impedance (Zhc). The seventh

Figure 8: s ymmetric harmonic power phas ors for h=5, 7 and 13

Cigre Science & Engineering • N°13 February 2019

111
Figure 9: long duration s imultaneous meas urements of harmonic voltage and current h=5, 7 and 13

Cigre Science & Engineering • N°13 February 2019

112
 Figure 10: alternative interpretation figure s hape
harmonic current is essentially made of current being
absorbed from the grid into the seventh harmonic filter
of the installation.
2. Based on figure 6 the harmonic current of the fifth, 13th,
and 19th harmonic order and the resulting voltage are
due to the combined influence of the background level
and the distorting installation under consideration
without any prevalence of one or the other. In Figure 9,
the points of the fifth, 13th, and 19th harmonic order are
not more or less aligned along the four straight lines
with a slope approximately equal to modulus of the
filter impedance (Zhc).
3. An alternatively interpretation of the 19th order cases
of Figure 9, is that the harmonic voltage and current
magnitude are low to deliver an accurate measurement
with little uncertainty: the measurement is less than
1% of Upr. This means that the methods of figure 6, and
8 of [1] could not applicable due to high measurement
uncertainty. The symmetrical voltages are below 1%
of Upr. In this range the relative uncertainty is rapidly
increasing when the magnitude of the voltage is
decreasing in the range Uh/Upr less than 1%, see dark
area Figure 4 .
An alternative interpretation needs to be investigated
given the diagram shapes of harmonic order 13th of
Figure 9.
The diagram shape of Figure 10 is not given in [1]. It might
be that the magnitude and angle of the red squares are
related to the harmonic background grid voltages minus
the harmonic current at the POC, and the combined grid
and distorting installation impedance. The assumption is
that the distorting installation harmonic impedance is lower
than the harmonic grid impedance. This means that the
angle of the red squares has a dominant harmonic distorting
installation impedance slope. Further notice should be
Cigre Science & Engineering • N°13 February 2019
113
given to the 13th harmonic order diagrams of Figure 8. It
suggests that the injected active current of the installation
is accompanied by absorption of reactive current of the
distorting installation. This consumption of reactive power
of the installation might compensate for the effects of
the injected active current of the distorting installation.
It might be that a harmonic elimination method of the
current source convertor has these effects. The calculated
distorting installation impedance is depicted in magenta
coloured squares. It might be that de cyan coloured line is
a measure for the harmonic grid impedance.
5. Conclusions
A practical analysis of a distorting installation on an
extra high voltages grid is performed. In this particular
case when the disturbance originates from the grid, an
assessment can be performed with known filter states
and calculated filter impedances. Based on [1] the
measurement shows that the 7th harmonic has its origin
in the grid. In case of the other harmonic orders it cannot
be concluded that, either the disturbance is due to the
grid, or due to the distorting installation, or both.
In the other cases the measured values where in the
range with a high level of measurement uncertainty. An
approximation of the relative combined measurement
uncertainty is made for the measurement system that is
used.
This paper considers the power directions of the
harmonics. This direction is related to the interaction
between the grid and the installation at the point of
connection. The interactions such as resonances and
cross-amplification of harmonics are not considered in
this paper.
6. Bibliography
[1] Working Group C4.109 CIGRE/CIRED. “Review of Disturbance
Emission Assessment Techniques” (CIGRE 468 June 2011)
[2] IEC/TR 61000-3-6 . “Electromagnetic compatibility (EMC) Part 3-6:
Limits Assessment of emission limits for the connection of distorting
installations to MV, HV and EHV power systems” (IEC 2008).
[3] IEC 61869-1:2007. “Instrument transformers Part 1: General
requirements” (IEC 2007)
[4] IEC 61869-2:2012. “Instrument transformers – Part 2: Additional
requirements for current transformers” (IEC 2012)
[5] IEC 61000-4-30:2008. “ Electromagnetic compatibility (EMC)
Part 4-30: Testing and measurement techniques Power quality
measurement methods” (IEC 2008)
[6] JCGM 100:2008. “Evaluation of measurement data — Guide to
the expression of uncertainty in measurement, JCGM member
Cigre Science & Engineering • N°13 February 2019
114
organizations (BIPM, IEC, IFCC,ILAC, ISO, IUPAC, IUPAP and
OIML) ” ( www.bipm.org 2008)
[7] IEC 61869-3:2011. “ Instrument transformers Part 3: Additional
requirements for inductive voltage transformers” ( IEC 2011)
[8] H. Yang. “HARMONIC EMISSION LEVELS OF INDUSTRIAL
LOADS STATISTICAL ASSESSMENT ” (CIGRE 1996 : 36-306)
[9] R. ROSS. “Harmonic Distortion Assessment Related to the
Connection of Wind Parks to the Brazilian Transmission Grid ”
(CIGRE 2016 C4-101)
[10] IEC 61000-4-7:2002+A1:2008. “Electromagnetic compatibility
(EMC) – Part 4-7: Testing and measurement techniques – General
guide on harmonics and interharmonics measurements and
instrumentation, for power supply systems and equipment connected
thereto” (IEC 2008)
[11] IEC TR 61869-103:2012. “Instrument transformers – The use of
instrument transformers for power quality” (IEC 2012)
Smart TSO-DSO interaction schemes
and ICT solutions for the integration
of ancillary services
from distributed generation
G. GUIDA, G. BRUNO, L. ORTOLANO, M. POLI, M. PALLESCHI Terna Rete Italia,
G. MIGLIAVACCA, D. MONETA, RSE,
C. ARRIGONI, F. ZANELLINI, Siemens S.p.A.,
G. DELLA CROCE, Selta,
A. BRIDI, M. BALDINI, Edyna
Italy

Summary and test in field technological devices to demonstrate that

The energy power systems are facing major challenges
as fossil fuel generation is replaced with renewable
generation, which is often characterised by variable
behaviour. This increases the need for resources to be used
to guarantee voltage and frequency stability, to deal with
congestion management and to ensure the compliance
with power quality standards. At the same time, there is
an increasing number of local, small-sized generation,
flexible demand and storage systems that are often located
at distribution level. These resources could potentially be
able to provide network services if they are aggregated
effectively. However, to achieve this, the roles of the
diverse network stakeholders – transmission systems
operators (TSOs), distribution systems operators (DSOs)
and aggregators – should be reshaped. In tandem with this,
the way real-time electricity markets are organised also
needs to be adapted to reflect the new operating paradigm.
The SmartNet project (http://smartnet-project.eu/) aims to
investigate different solutions to optimise the coordination
between TSOs and DSOs with the purpose of improving
the grid monitoring and the participation of resource
located in the distribution grid in ancillary services.
Within this project, three physical pilots are implemented
in Italy, Denmark and Spain to determine new practical
solutions.
The subject of this paper is the Italian pilot, realized by a
consortium composed by RSE, Terna (TSO), local DSO
Edyna, Siemens and Selta in order to develop, implement
a smart management of the entire electrical grid, with the
coordination of all the actors involved, could help to face
new challenges.
In particular, the interaction and the collaboration between
TSO and DSO could improve the observability of the grid
and allow investigating and defining needs, opportunities
and feasibility to employ the potential contribution from
all the resources connected in the transmission and
distribution grid. In this way, the electrical system could
be more flexible and more optimizable.
In terms of technology, the Italian pilot project aims
to implement new devices in order to improve the
observability of the DSO’s grid and to enable flexible
generation to provide system services. In particular, the
main functionalities regard:
- The acquisition of real time information about the
operational state of the power plants located at
distribution level
- The estimation of the production fed into the power
grid by unmonitored plants
- The aggregation of real time data at the
interconnection point MV/HV provided by the DSO
to the TSO
- The voltage regulation by generators connected to
HV and MV levels
- The power/frequency regulation (Automatic
Frequency Restoration Reserve – aFRR) by
generators connected to MV grid

*luca.ortolano@terna.it

KEYWORDS
Ancillary services markets, Observability, Voltage regulation, frequency/power regulation, distributed generation,
DER, RES

Cigre Science & Engineering • N°13 February 2019

115
The purpose of the present paper is to submit the
achievements of the demonstration project realized in
Italy during the first two years and to describe the ICT
solutions developed by Siemens and Selta in order to
fulfill the above functionalities.
1. Introduction
The Italian energy context, as the European one, is
undergoing important changes. During the recent years
there has been a rapid increase in the number of renewable
energy sources also due to government incentives: 5,6
GW of wind power capacity and 18,5 GW of solar power
capacity has been installed since 2008
An important consequence is the spread of generation units,
in particular photovoltaic panels, located at distribution
level. The energy framework is moving from a power park
module characterized by few traditional plants connected
to HV grid to an energy context composed by numerous
small-sized generators spread at MV and LV levels. Only
Italian photovoltaic installations amount to 50 traditional
plants of 320 MW each, but located at distribution level.
This new configuration of the network can lead to reverse
power flows: in case of local oversupply, when the amount
of locally generated power exceed the local load, the
active power can rise up from distribution grid to upper
grid level. In the future, it could influence the voltage and
frequency regulation also in the transmission grid.
Furthermore, due to the unpredictable nature of the
renewable energy resources, a future increase of renewable
penetration will lead the needs to improve the infrastructure
for monitoring and control of MV and LV levels.
Within this context, the Italian pilot project of SmartNet
aims to demonstrate the advantages of improving the
observability of the grid and using the flexibility from
Distribute Energy Resources (DER) and Renewable
Energy Sources (RES) in the ancillary services; in the
future both transmission system operators and distribution
system operators could benefit from these types of
flexibility.
With these opportunities, the technological partners,
Siemens and Selta, have developed devices that are able
to implement the functionalities necessary for the scope.
Siemens has realized the HVRS (High Voltage Regulation
Cigre Science & Engineering • N°13 February 2019
116
System) that implements the algorithms to perform the
function of the reactive power information aggregation
and to control the high voltage generation to achieve the
voltage regulation.
Selta and Siemens have realized, with two different
ICT solutions, the MVRS (Medium Voltage Regulation
System), which aims to perform information aggregation
and also to control the MV generation to achieve both the
power/frequency and the voltage regulation.
This paper illustrates in details the architectures and the
functionalities implemented.
2. The grid
In the following, the grid under investigation, where the
SmartNet project has been realized, is briefly descripted.
The pilot is located in the area of Ahrntal valley, in South-
Tyrol, alpine Italian region, at the border with the Austria
(Figure 1 – left side).
The orography and the climate of this region has always
been favourable for the exploitation of water power to
product energy: the abundant winter snowfalls give rise to
abundant and numerous water flows during the summer.
The choice of this location is due to the availability of
several generating modules of different size connected to
all levels of voltage.
On one hand, regarding the HV part of the project, the
subject of study is the TSO HV substation, where due
hyroelectric plants are directly connected to TSO’s grid
(Molini and Lappago of 43 MW total).
On the other hand, part of the project involves the DSO
grid. As indicated in Figure 1 (right side), the MV grid of
this area is powered by the primary substation “Molini
di Tures” of EDYNA characterized by 2 transformers
HV/MV (132/20 kV) of 40 MVA each. There are 9 MV
feeders outgoing from this substation: 3 of them feed
distributed load at LV level (mainly domestic consumers,
toursitic load and small-sized industry) and the other 6 are
dedicated to MV producers or local DSOs.
For the characteristics of the area, these feeders are very
long and without connections with others feeders; the
same substation Molini di Tures has only one possible MV
connection with another HV/MV substation (Brunico).
 Figure 1 The Arnhtal valley in Europe (left side) and EDYNA MV grid of Molini di Tures (right side)
At the MV grid, there are connected 23 producers, with
an installed power of 29 MW (27.7 flowing hydroelectric,
1.5 biothermic, 0.2 PV) and 5 local DSOs that are
characterized by a small number of costumers fed by
one or more hydroelectric plants. For this reason, these
interconnection points with DSO grid are comparable to
big prosumers with 17 MW of total power consumption
in the interconnection points. There are also 0.85 MW of
production in LV (0.73 PV).
The consequence of this installed production at MV
and LV levels is that for almost the whole year in the
interconnection point between TSO and DSO the power
rise from MV to HV, with a peak higher than 30 MW in
summer, as depicted in Figure 2 This characteristic makes
this grid suitable to implement the project.
In order to realize the project, Selta has installed the
Plant Central Regulator (PCR): all the 28 interconnection
points are monitored by this device and 7 of the biggest
hydroelectric plants, of 22 MW total, are also controlled
by the PCRs to obtain the voltage and the frequency
regulation.
PCR is an integral part of Selta’s solution for DSO and it
Figure 2 Average load in Molini di Tures (year 2017)
Cigre Science & Engineering • N°13 February 2019
117
represents the most peripheral device in the communication
chain. It plays an essential role for observability of the
DSO grid, for the Voltage
Regulation and power/frequency Regulation purposes.
It allows the monitoring and the control functions of the
MV DERs. Indeed, when PCR interfaces a controllable
DER, it makes available the functions of reactive power
and cosphi modulation in V Regulation and active power
modulation in the P/f Regulation. PCR can offer protocol
or physical I/O interface modalities, according to the
equipment of the interconnected plant. Furthermore PCR
is able to adapt itself to plant configuration and, at the same
time, it allows to monitor measurements from generators,
storage systems, adjustable loads and from connection
points on the grid. Its architecture integrates also 3G/4G,
Distribution Line Carrier (DLC) and optical fibre
technologies, leading to a safe and reliable communication
service. PCR is considered the prototype of Italian CCI
(Controllore Centrale d’Impianto), promoted and required
by CEI 0-16 to control and to measure DERs in MV grid.
Furthermore it has already been tested in other practical
applications and experiments. [12]
3. Observability
The widespread presence of non-programmable energy
sources creates the opportunity to develop innovative
strategies to increase the monitoring and the control of
the grid. With this purpose, the Italian NRA’s (ARERA)
resolution 646/2015/R/eel [2] aims to promote the
implementation of services to allow a reliable and effective
management of distribution grids characterised by a high
penetration of Distributed Generation (DG). It defines
incentives in favour of DSOs to stimulate the investments
focused on the observability of the DG (power flows
and states). Within this resolution, the Authority has
defined two types of MV observability functionalities,
characterized by different level of complexity: OSS-
1 to send to the TSO data e measures of the renewable
generators and OSS-
2, which requires a more complex architecture, to send
also the estimated values of generation and load.
An important aspect of the observability is the aggregation
function that consists of obtaining equivalents of the
generation and the load on the HV/MV transformer: the
real time information at the interconnection point between
TSO and DSO are collected and aggregated by the DSO.
The nodal equivalentation allows the TSO to have a better
observability and transparent access to the underlying
resources to ensure a safe use of this type of resources. In
fact the possibility for the TSO to know the total amount
of the generation, by type of energy source, and of the
load at the HV/MV substation at all times allow the TSO
to have:
- Better perception of the production location and the
actual load supplied
- Better knowledge of the actual absorption to restore
the service after the disconnection at the station
- Better assess the N-1 security of the grid
- Better realize dynamic evaluation to individuate
critical constrains
- Better manage the Defence System
The data transmitted by the DSO to the TSO are:
- Total generation power installed differentiated by
source, i.e. PV, rotating machine etc.
- Total load, understood as gross amount of load
compensated by distributed generation
- Real time values of the active and reactive power
Cigre Science & Engineering • N°13 February 2019
118
flows every 20 seconds, differentiated by load and
generation source.
The first goal of the Italian pilot is the development of the
functionality to obtain an accurate observation of the MV
an LV sources real time.
MVRS represents the application software that,
together with SCADA system, constitutes the Advanced
Distribution Management System (ADMS) solution for
Smart Grids. Therefore, MVRS is considered as added
value for a traditional SCADA system, especially for
electrical grid characterised by the presence of Distribute
Energy Resources (DERs). A key point of the overall
system is that MVRS can manage whatever kind of
energy source (renewable energy resources, storage
systems, adjustable loads) and it is extremely flexible
to the grid configuration. ADMS obtains real time field
information via multiple stack interactions (IEC-61860,
IEC 6870-5-104, Web Services, IOT systems). In this
innovative solution, the communication between MVRS
and SCADA takes place through tele-control protocols
and inter-process communication solutions, by allowing
technological distributed architectures.
Acquisition of measurements, collection of nominal data,
estimation about DERs power (now cast and forecast),
grid reconfiguration and aggregation of information are
the main tasks of the observability module, implemented
by real time processes. Powers, voltages and weather data
are just some examples of exchanged variables used for
the observability of the Distribution grid.
Aggregation of real time information is based on
power measurements from the field along with power
estimations. The final result wants to be the creation of
power aggregates, divided according to energy source,
referring to distributed resources available in the part of
grid under examination. For the purpose of responding
to ARERA resolution 646/2015/R/eel, the aggregation is
composed and sent to the TSO every 20 seconds.
Estimation of power generation is necessary for the
unmonitored MV DERs and for the whole LV production,
whose measure is usually not available in real time for
technical and economic reasons. Because of nature of
renewable energy sources, appropriate methods have to
been carried out for power estimation.
In the following, we will describe in detail the two
different approaches developed by Siemens and Selta in
order to compute estimations based on a limited set of
measurements on the network.
The Selta now-cast estimation algorithm is based on
the concept of Sentinel Measurements (SMs). The SMs
can consist of real time power measurements related to
plants monitored by PCRs, of historical power profiles
acquired by energy meters or of real time weather data
(e.g. solar radiation, air temperature or river flow). Taking
into account the historical data, it’s possible to calculate
the correlation coefficients between the unmonitored
DER and the SMs. This off-line evaluation allows to find
out the relationship between neighbouring plants of the
same energy resources (e.g. hydroelectric generators) and
to assess the impact of weather variables (in particular
for the PV plants). Therefore, the best SMs (3 SMs in
practical applications) are selected for each unmonitored
DER and appropriate weights are generated.
In order to follow the seasonality of the power generation,
for each month the set of SMs is updated and the correlation
coefficients are evaluated. The real time estimation of
DER power amount is based on the weighted average of
the selected SMs, correlated by the respective coefficient.
A different approach is carried out as regards the estimation
of reactive power exchanged by the unmonitored plants.
Considering the historical metering, the average value
of cosphi is calculated for every generators. Therefore,
the on-line evaluation of reactive power is computed as
directly related to the estimation of active power through
the corresponding cosphi.
The Siemens solution is based on an integrated architecture
that is able to elaborate and combine non- homogenous raw
data. To estimate the real and reactive power amount for
each unmonitored MV plant and each MV/LV Secondary
Substation transformer, the system is able to consider real-
time weather data (solar radiation and temperature), “near
real-time” data registered by energy meters for MV plants
(on 15 minutes basis, originally used for billing purposes)
and real-time data acquired by PCR.
The weather data are processed by a specific Siemens
application based on a self-learning algorithm that,
considering the geographical location and the physical
characteristics of each plant, is able to provide the so
called “nowcast” for all the PV plants connected to the
Cigre Science & Engineering • N°13 February 2019
119
DSO electrical network (MV plants and LV generation
connected to each MV/LV transformer). All the other
energy resources (i.e. non-renewable or hydroelectric) are
estimated processing the combination of historical data
and available “near real-time” data registered by energy
meters. All the “nowcast” data is then transferred to the
Siemens monitoring&control system that carries out the
selection and combination of this data with the real-time
measurements coming from the field (PCR) to obtain the
aggregated real and reactive power (HV/MV transformer
level) ready to be sent to the TSO.
In order to evaluate the aggregation functionality and
assess the accuracy of the algorithms implemented, an
ex post analysis evaluates the Estimation Precision Index
(EPI): it relies on the comparison of monthly metering and
the corresponding estimation. In particular, it’s subject of
study the determination of the amount of measures of
monitored DER necessary in order to obtain the requested
accuracy
4. Voltage regulation
Another consequence of the growth of the importance
of the renewable generation and of the small- size plants
connected to distribution grids is the necessity of an
improvement of the reactive power management.
The ARERA resolution 300/2017/r/eel [3] is an important
step in this direction because it proposes to realize pilot
project about modalities for the remuneration of ancillary
services currently non- remunerated as the voltage
regulation. The aim is to collect useful elements for the
reform of the dispatching market.
For this reason, the project aims to preempt future
evolution of the regulatory framework, with a view to a
possible opening of the reactive power market, realizing
devices able to modify the reactive production/absorption
of the users controlled.
In detail, on one side, Siemens’ HVRS has been installed
in Terna’s substation to control the reactive power of
hydroelectric plants connected to HV grid, which currently
are only involved in the primary voltage regulation.
On the other side, MVRS devices perform similar
functions to manage the reactive power contribution of
distributed plants at MV level; the purpose is to run the
Primary Substation like a virtual generation plant able to
 Figure 3 Proposed substation voltage control structure (left side) and control law (right side)
exchange data and interact with the TSO like a traditional
HV plant (VVP Virtual Power Plant). On the basis of the
grid configuration and plant’s conditions, they compute
the virtual capability to determine the reactive availability
and exercise control actions in order to comply the TSO’s
command.
4.1. HVRS voltage regulation [11]
The evolution of electric system scenario introduced above
will benefice by the involvement of non- programmable
renewable generating facilities in the operation of electric
power system to maintain and, where possible, to increase
the margins of security and reliability of transmission and
distribution systems.
Today, only conventional large power plants are equipped
with specific devices to participate at the hierarchical
control voltage and one of the goals of the SmartNet
project is to implement and test an innovative tool to
enable the involvement of RES in dispatching function,
such as photovoltaic plants, wind farms and small
hydroelectric plants.
In SmarNet application, the HVRS device aims to smooth
the voltage fluctuations measured in the 132 kV Molini
di Tures substation by means of the reactive power
exchanged by the two hydro power plants “Molini” and
“Lappago”, connected at this substation (two synchronous
generators for each power plant). The HVRS cubicle
is designed as a SAS (Substation Automation System)
functional unit [10], so enabling an easy integration in
the TSO substation automation architecture. The HVRS
algorithm measures the HV busbar voltage and real power
generation of the hydro plants and coordinates the reactive
resources directly connected according to a control law
that links the voltage error (namely the difference between
the measured and optimal voltage) and generator reactive
power contribution, (reactive power injected/absorbed,
power factor or tanphi, as depicted in Figure 3.a, left side).
The calculated set point is compliant with the generators
capability charts, parametrized in the HVRS database and
Cigre Science & Engineering • N°13 February 2019
120
sent to the TSO. The busbar voltage/power factor control
law must be robust with respect to all different operating
points and resilient against contingencies and large
perturbations: it represents an evolution of the regulation
algorithm proposed in [11]. The proposed control law is
depicted in Figure 3.b (right side): a continuous control
centered in the voltage reference optimal set point (the
sample value, 1.02 pu, can be remotely updated by the
Terna EMS) and a cubic correlation between the voltage
error and generators power factor is adopted. Such a
control law can be changed parametrically: Figure 3.b
depicts three different control laws, increasing the control
effort moving from the violet to the blue one. In particular,
the cubic correlation has been introduced to obtain a
flexible control of the reactive resources.
The HVRS send to each control system of the generators
the reactive power set-point which is implemented by
varying the voltage set-point of the generator in order to
obtain the request amount of reactive power. The reactive
power control loop is slower than the voltage control loop,
and this allow to guarantee a fast response of the generator
in case of sudden change in the network operation (e.g.
short circuit).
The HVRS track the voltage error (the difference
between the voltage set-point and voltage measurement);
at first instance the generators will react to the voltage
error and then the HVRS will send a reactive power
setpoint proportional to the voltage error following the
law described in the diagram figure 3.b. The change of
reactive power exchange of Molini di Tures and Lappago
power plants make the error smaller and then a smaller
requirement of reactive power variation will be send to
the generators control systems of the power plants. If the
needed reactive power exceed the available amount the
request will not be augmented anymore. If it is not the
case then the HVRS will allow with intermediate steps to
reach the HV voltage setpoint with accuracy less or equal
to the voltage error deadband.
 Figure 4 Main HMI of HVRS device
n a higher level, the TSO could adopt the proposed
architecture to control the subtransmission network; as
a matter of fact, the TSO should define, for each SAS,
a voltage set point and a regulation “strength” (i.e., a
control law) in order to coordinate a large number of RES
generators by means of few signals. In operational terms,
on the basis of the operational state of the plants, the
HVRS is able to supply to the TSO the current availability
of reactive power and to split the command amongst the
controlled generators in order to obtain a homogenous
division of the efforts and avoid undue reactive power
flows between providers in the same electrical area.
In Figure 4 the HVRS main HMI is represented: in the
upper part the regulation parameters are summarized,
while the orange boxes resume the regulation operating
status and the calculated set point. The single line diagram
of Molini di Tures substation is then reported, together
with the active/reactive power and voltage measures.
4.2. MVRS voltage regulation
In order to further increase the amount of reactive power
reserve, for the benefit of DSO and/or TSO, SmartNet
project aims also to involve DG in the voltage regulation
ancillary service.
With this purpose, the second functionality implemented
in the MVRS is the voltage regulation through units
connected at distribution level which define the Virtual
Power Plant (VPP): the task of the Advanced DMS is
to realize a control loop to fulfill the TSO set points on
the TSO-DSO interconnection point (HV busbar voltage
set point or reactive power set point). The entire control
Cigre Science & Engineering • N°13 February 2019
121
process takes into account primarily all the technical
constraints on the DSO network: in fact it defines the
reactive power dynamic capability considering under and
over excitation limits dictated by DSO’s grid constrains
and then it realizes the TSO set points in order to solve/
avoid violations on the MV network. For these purposes,
the Advanced DMS considers both the regulation of the
HV/MV transformer On Load Tap Changer (OLTC) and
the DERs reactive power.
In this paragraph, two technical solutions, implemented
by Siemens and Selta, will be described.
Voltage Regulation, implemented by Selta, works with a
continuous time constant cycle and it consists in a multi-
stage algorithm. First step is characterized by the state
estimation of the electrical parameters of the Distribution
grid model, corresponding to the controlled portion.
Measurements from SCADA, power estimations and
load flow calculation contribute to the smart allocation
of unmonitored loads, feeder by feeder. The on-line
evaluation allows to check out the nodal voltages value
at any time. Considering a normal operating range in
±5% of nominal voltage, second step includes the voltage
regulation by OLTC and/or by DERs reactive power in
order to fix possible nodal voltage violations in the MV
grid. This preliminary regulation takes into account the
DERs’ sensitivity to the node in violation and it tries to
avoid the misuse of OLTC. The specific calculated set
points are sent to the involved controlled resources. Until
the violation persists, VPP is declared unavailable to
participate in the ancillary services requested from TSO.
If program does not detect any violations, the evaluation
of VPP reactive power dynamic capability represents the
next step. In this part of the algorithm, a series of load
flow are run in order to evaluate the maximum available
capability of the VPP in under and over excitation. The
calculation includes checking of possible nodal voltage
violations; where necessary the capability is reduced
by DERs sensitivity method. Reactive power virtual
capability and operation point at the HV side of the
transformer are sent to TSO, on the basis of set up and
topology of the DSO network.
According to VPP capability, the TSO requires ancillary
services to regulate high side voltage value at TSO/DSO
grid interconnection point. MVRS can receive a percentage
of reactive power referred to the last capability sent. In
other cases, voltage set points are received and appropriate
Q(ΔV) tables convert them in reactive power ones.
The goal of the regulation algorithm is chasing and keeping
set point requested, adapting decisions to changing field
conditions. Therefore, a smart splitting of reactive power
set point is implemented according to single DERs
capability. The MVRS closed loop control system finds
a solution for the simulated model and the set points are
sent towards MV DERs, available and controllable. This
regulation step is repeated only after the real settling of
the plants, working as external closed loop control.
The Siemens MVRS implements a control loop based
on the evolution of some functionalities of the Siemens
Smart Distribution Management System [4] [5]; in detail,
to realize this SmartNet functionality there are three
algorithms that provide the MV network State Estimation,
the reactive power Virtual Capability and the Voltage
Regulation.
In case of receiving from the TSO a voltage set point (for the
Primary Substation HV busbar), the MVRS pre-processes
the data using specific Q(ΔV) curves (implementing the
same approach used in the Siemens HVRS); thanks to this
elaboration, the Voltage Regulation algorithms are always
run with a reactive power constraint assigned to each HV/
MV transformer.
Regarding the State Estimation, this process represents the
first step to manage the Distribution Network in real time,
before taking any control action. In Distribution Networks
the application of classical state estimation methods is
Cigre Science & Engineering • N°13 February 2019
122
impossible mainly due to the reduced number of available
measurements and, hence, the lack of redundancy.
Therefore, depending on the available measurement points
various specially designed State Estimation techniques
have been implemented: (i) simplified algorithms when
only the primary substation measurements are available
[6][7] and (ii) advanced algorithm when both primary and
secondary substation measurements are available [8].
About the control step (Virtual Capability computation
and Voltage Regulation), this part of the functionality
aims to apply an imposed reactive power set point on
the HV–MV interface (the HV– MV transformer) by
optimally dispatching the reactive resources available at
the MV level and assuring the operational constraints of
the network.
Originally, an optimization problem was designed by
adapting the traditional Optimal Reactive Power Flow
(ORPF) problem [9] to an ORPF model suitable for
Distribution Networks [5]. Here, the goal of the ORPF
was to minimize the real energy losses in the Distribution
Network subject to the following constraints: (i) power
flow equations; (ii) generating units capability constraints;
(iii) nodal voltage operating limits and (iv) branch current
limits. To realize this specific SmartNet goal, the objective
function of the ORPF [5] was changed to minimizing the
square of the difference between the actual reactive power
flowing in the HV–MV interface and its set point; ideally,
this is null, but if constraints are at limit then the minimum
possible deviation from the set point is achieved.
With this new algorithm, it is possible to: (i) calculate,
at a given moment in time, the reactive power limits at
the HV–MV interface by giving fictitious large set points
(above the sum of the capabilities of the generating units)
for both capacitive and inductive operating conditions
(i.e. Virtual Capability computation) and (ii) impose, in
real time, a reactive power set point that lies inside the
previously computed limits (i.e. Voltage Regulation).
5. Frequency/power regulation
The last functionality implemented in the MVRS is
the frequency/power regulation through DG. Since the
Authority, within the resolution 300/2017/R/eel, has open
the dispatching market to aggregation of non-relevant
units and of loads, this innovative implementation has
become a very topical issue. Regarding this functionality,
the main task of the MVRS is to receive a level command
by the TSO and act an active power variation of the
VPP in order to modulate active power fed into the grid
complying with the TSO requirement.
In practice, the MVRS, considering all the available data
(topology, measurements, etc.) characterizing the MV
electrical network, calculates an aggregated dynamic real
power capability (i.e. the real power range available for
the TSO to provide the power/frequency regulation).
Each MV generator included among the plants that are
available for this kind of regulation, is characterised by
two real power regulation ranges: one for real power
increasing and one for real power decreasing. The
preliminary computing on these data are performed in
different ways by the two MVRS: in Siemens solution,
the real time regulation ranges for each MV plant are
pre-processed considering equivalent generators (the
individual physical generators are clustered considering
homogenous primary energy source and nominal power);
in Selta approach, the pre-processing is done considering
the actual state of each switch available in the plant for
the single physical generators. After this step, the MVRS
aggregates the single real power regulation ranges to
describe the virtual generator (i.e. the Primary Substation).
The activation of the power/frequency regulation starts an
open loop real power regulation where the set point sent
by the TSO (0-100 % format) has to be processed by the
MVRS firstly to calculate the necessary total real power
and then to split it according to each unit capability. The
final real power set points are sent to each PCR installed
on field to actuate the power regulation. As required by
the TSO, the entire control loop considers very strict
time requirements: provide to the TSO actual regulation
ranges with a 4 seconds schedule and process/send to
the field the set points with a 8 seconds schedule; from
this point of view, besides the data processing speed, the
communication channel between the TSO/DSO and to the
PCR on field plays a fundamental role.
6. Conclusion
The paper presented in detail SmartNet pilot project
realized in Italy, highlighting the advantages of an
integration of renewable energy sources and distributed
generation into the system. In particular, the devices
developed by Siemens and Selta allow to obtain an
Cigre Science & Engineering • N°13 February 2019
123
improved monitoring of the grid and to exploit the
potential of the resource that could not participate in
system services until recently.
As shown above, the regulatory framework is taking in
account the evolution of the energy system and it is moving
in this direction in order to encourage the development
of a smart management of the grid. In this context, the
project represents an innovative study that will provide
important information, also through the forthcoming
execution phase, in order to collect elements to allow in
the future the improvement of the grid management for
the benefit of the TSO and the DSO.
7. Bibliography
[1] G. Migliavacca, H. Gerard, M. Džamarija, S. Horsmanheimo, C.
Madina, I. Kockar, G. Leclerq, M. Marroquin, H. Svendsen, “TSO-
DSO coordination and market architectures for an integrated
ancillary services acquisition: the view of the SmartNet project”, 2018
Cigre Session
[2] ARERA resolution 646/2015/R/eel
[3] ARERA resolution 300/2017/R/eel
[4] A. Berizzi et al., “Architecture and functionalities of a smart
Distribution Management System,” 2014 16th International
Conference on Harmonics and Quality of Power (ICHQP), Bucharest,
2014, pp. 439-443.
[5] C. Arrigoni et al., “Smart Distribution Management System:
Evolution of MV grids supervision & control systems,” 2016 AEIT
International Annual Conference (AEIT), Capri, 2016, pp. 1-6. [6]
A. Berizzi et al., “Advanced functions for DSOs control center,” 2013
IEEE Grenoble Conference, Grenoble, 2013, pp. 1-6.
[7] C. Bovo, V. Ilea, M. Subasic, I. Rochira and C. Arrigoni, “Innovative
methodology for observability improvement of distribution networks,”
2015 IEEE Eindhoven PowerTech, Eindhoven, 2015, pp. 1-6.
[8] Bovo C., Ilea V., Subasic M., Zanellini F., Arrigoni C., Bonera R.,
“Improvement of observability in poorly measured distribution
networks”, Proceedings - 2014 Power Systems Computation
Conference, PSCC 2014.
[9] V. Ilea, C. Bovo, M. Merlo, A. Berizzi and M. Eremia, “Reactive power
flow optimization in the presence of Secondary Voltage Control,” 2009
IEEE Bucharest PowerTech, Bucharest, 2009, pp. 1-8.
[10] A. Berizzi, C. Bovo, J. Allahdadian, V. Ilea, M. Merlo, A. Miotti, F.
Zanellini, “Innovative automation functions at a substation level to
increase RES penetration”, CIGRE 2011 Bologna Symposium - The
Electric Power System of the Future: Integrating Supergrids and
Microgrids, Bologna, 2011.
[11] M. Petrini, C. Sabelli, A. Miotti, F. Zanellini, S. Koeppe, “ A new
substationi level dispatchin approach to increase the network control
and perfomrances in presence of dispersed generation”, 2012 Cigre
Session
[12] M. Pareti., “Coinvolgimento della Generazione Distribuita nella
qualità e stabilità della fornitura elettrica”, ANIE – Forum del
Telecontrollo, Verona, 2017.
Effectiveness of different flexibility options
and innovative network technologies for
the use in the BDEW traffic light concept,
on the basis of a German distribution grid
T. Ackermann*, E. Tröster, L. Hülsmann, Energynautics GmbH, Germany
U. Ohl, M. Koch, EWR Netz GmbH, Germany
Summary
The continuous expansion of distributed generation (DG),
mainly rooftop and free-field photovoltaic (PV) systems
as well as wind power plants, is changing the power
flow landscape in the distribution grid: During times of
high solar irradiation and wind speeds, generation can
exceed load on a distribution feeder by a multiple. This
pushes cables and transformers to their thermal limits
and may lead to voltage deviations outside of the allowed
+/-10 % of nominal value. Furthermore, new electrical
consumers, such as electric vehicles and heat pumps, may
also aggravate the network during peak demand times.
However, these situations occur only during few times of
the year which makes traditional grid reinforcements an
expensive countermeasure. Instead, shifting demand from
times of high load to times of high PV and wind power
generation can offer cheaper alternatives to mitigate grid
congestion and delay or prevent grid reinforcements. In
light of this, the German Association of Energy and Water
Industries (BDEW) has developed the “BDEW smart grid
traffic light” concept. Its aim is to communicate regional
flexibility needs during the amber phase by means of a
local flexibility market in order to stay in the green phase,
as opposed to the red phase where a grid congestion
prevails. DESIGNETZ, one of five showcases of the
German funding program “Smart Energy Showcases -
Digital Agenda for the Energy Transition” (SINTEG), is
intended to demonstrate the working principles of such a
flexibility market. This paper deals with the simulation
results of the traffic light concept. For this, a radial medium
voltage (MV) feeder in a rural area is analysed, supplying
multiple small villages. The feeder is heavily penetrated
* t.ackermann@energynautics.com
KEYWORDS
Grid congestion - Overvoltage - Distribution grid - Flexibility market - Demand side management - Electric vehicles -
Heat pumps - Traffic light - BDEW
Cigre Science & Engineering • N°13 February 2019
124
by PV generation but still only 22% of suitable rooftop
area is occupied by PV. The PV generation has been
further increased and, additionally, new electric vehicles
and heat pumps were modelled, to represent a scenario
of the feeder in the year 2030. Load flow calculations in
DIgSILENT PowerFactory show that this leads to voltage
violations outside the defined ranges. Three flexibility
options, namely shift in electric vehicle and heat pump
demand as well as PV curtailment, are utilized in the
traffic light concept and chosen based on their sensitivity
to alleviate the grid congestion as well as their costs. The
simulation results show that the concept can effectively
allocate the most cost- effective flexibility options and
keep the MV feeder within its safe operation boundaries.
This lowers PV curtailment by 15 % for the analysed
grid and reduces costs by 6 % compared to the case
where only PV curtailment is applied. The cost benefits
of such a concept need however be weighed against the
communicational and regulatory challenges that such a
flexibility market poses.
1. Introduction
With increasing capacities of distributed renewable
generation, such as photovoltaics (PV), being connected
to the distribution networks, distribution system operators
(DSOs) are facing a number of challenges. In grids
that were designed for unidirectional flows from the
transmission grid to the end customers, high shares of
distributed generation can lead to voltage and overloading
problems (grid congestion). Voltages have to be kept
within a +/-10 % band of the rated voltage [1], while cables
and transformers need to stay within their thermal rating.
Congestion in the distribution grids may be aggravated
 Figure 1: Analysis of rooftop PV potential by GIS data analysis (top) and categorization according to orientation (bottom)
further by increasing numbers of heat pumps and electric
vehicles (EV) that increase electricity consumption. New
control strategies are thus necessary to alleviate local grid
congestion problems. Demand side management (DSM)
can be used to shift heat pump operation and EV charging
from times of high demand to times of low demand or
excess PV and wind power generation, increasing the
hosting capacity of a grid for both renewables and new
demand.
Providing this flexibility at the right location and at the
right time requires both IT infrastructure that enables
the necessary communication, and a defined regulatory
framework governing all market participants to work
effectively towards the three general energy policy goals:
Security of supply, affordability and sustainability.
In this sense, the German Association of Energy and Water
Industries (BDEW) has proposed the BDEW traffic light
system. It enables DSOs to signal a local flexibility market
if a grid congestion problem prevails. Subsequently, the
amount of necessary flexibility is determined in a specific
grid segment. Flexibility providers and aggregators then
compete against each other and the most cost-effective
providers are chosen.
This paper is part of the DESIGNETZ project [2], one
of five showcases under the German funding program
“Smart Energy Showcases – Digital Agenda for the
Energy Transition” (SINTEG). It seeks to test the BDEW
traffic light mechanisms through simulations on a model
of a real distribution network described in Chapter 2. The
BDEW traffic light system is explained in more detail
in Chapter 3, followed by a description of the modelled
flexibility options, namely heat pumps, EVs and PV
curtailment. Finally, simulation results are presented in
Chapter 5 and a conclusion given in Chapter 6.
In subsequent project phases the BDEW traffic light
system will be put into real-life practice and tested for its
practicability and effectiveness.
Cigre Science & Engineering • N°13 February 2019
125
2. Model setup
2.1. Distribution feeder model
To study the effects of demand and generation flexibility
embedded in a flexibility market framework, steady-state
load flow calculations are performed with DIgSILENT
PowerFactory on a 20 kV distribution feeder located close
to Worms, Germany. Only the MV network is modelled,
no LV grids are considered. Loads are predominantly
households, modelled with standard load profiles, and
aggregated through a total of 49 MV/LV substations. Line
types consist mostly NA2XS2Y 3x1x150 cable types and
Al_St_95 overhead line types. The feeder has a length of
about 22 km, with a large 7.3 MWpeak (MWp) PV power
plant located at a distance of 17.5 km from the primary
substation and only short side-branches. Together with an
installed PV capacity of 4.4 MWp in the LV grids, this is
already leading today to large voltage deviations of around
0.05 p.u. along the MV feeder, hence allowing only a
small voltage band for the LV grids in order not to violate
the EN50160 threshold of 1.10 p.u. Additional rooftop PV
generation as well as new heat pump and electric vehicle
demand are added, to represent shares expected for the
year 2030. They are described hereinafter.
2.2. PV development
Only rooftop PV potential is considered and no new
ground-mounted PV power plants were added. Rooftop
PV potential was estimated using OpenStreetMap
building and residential area data [3]. No building data
was available for some substations and an approximation
based on the residential area was used. The orientation
of each building was evaluated, by considering the
longest side of the building to be the dominant factor
to determine its roof’s orientation. Only south oriented
buildings, as shown in Figure 1 were considered suitable
for PV installations and only 50 % of the building area
could be used for PV systems. Approximately 60 % of
all buildings in the area were facing south. This led to a
potential for PV rooftop installations of 20 MWp, with 4.4
MWp rooftop PV already installed along the feeder. The
installed rooftop PV power was then increased by adding
PV plants at each substation until one of the following
thresholds of the total PV potential was reached at the
substation: 30 %, 50 %, 70 %. Anything above 70 % of
total PV potential was not increased. This results in a 68%
increase in rooftop PV capacity in 2030, corresponding to
a total of 7.4 MWp.
Figure 2: Electrical demand of the heat pump for heating, hot water
and in total
2.3. Heat pump model
For 2030, a heat pump penetration rate of 11.9 % according
to [4] is assumed. Heat pumps were modelled with an
aggregated profile of 40 heat pumps. This is roughly the
average number of heat pumps per village connected
to the feeder in 2030. Measured solar irradiation and
temperature values for the entire year 2015 in hourly
resolution, obtained from PVGIS [5], were used as inputs
for heating demand. Heating demand was disabled if
Cigre Science & Engineering • N°13 February 2019
126
outside temperatures exceeded 15 °C. To determine hot
water energy demand, the CREST demand model [6], a
high-resolution stochastic model of domestic thermal and
electricity demand, was used. The combined heating and
hot water energy demand, in a 15-minute resolution for a
whole year, was then scaled to the number of heat pumps.
Based on both heat demands for heating and hot water,
the total electricity consumption of the heat pumps can be
obtained according to (1).
(1)
Q: Heating and hot water demand
COP: Coefficient of performance of the heat pump
Tft: Flow temperature (heating 40 °C, hot water 60 °C)
To: Outside temperature
Normalizing the aggregated profile of 40 heat pumps to
a single heat pump results in the profile shown in Figure
2. It can be observed, that due to the aggregation, the
maximum value of the averaged electrical consumption of
the single heat pump is lower compared to the maximum
capability of typical heat pumps (e.g. 3 kWel). Further, a
seasonal trend can be observed in the total consumption,
which results from the significantly lowered electrical
demand for heating during summer as well as higher
outside temperatures that improve the COP of the heat
pump. The electrical demand for providing hot water is
also slightly lower in summer, mainly caused by higher
outside temperatures.
2.4. EV model
The EV modelling is based on a large-scale German
mobility study [7]. In this study, data has been collected
on trip numbers depending on weekday, time and activity
as well as driving duration. Based on these, probability
distributions are set up and a one-year driving profile is
randomly created for each EV, specifying departure and
arrival times as well as driving distance.
The driving profiles are then translated into a charging
pattern with the assumptions as seen in Table I. It is
assumed that the EV is recharged only if the SOC at
EV arrival has fallen below 50 % or if one of the next
two trips will fully deplete the battery, assuming perfect
foresight of the EV user’s upcoming travels. Charging the
EV as soon as it returns, keeps the average SOC high,
which has detrimental effects on battery lifetime [8], [9]. has proposed the BDEW traffic light concept. The idea
Additionally, this charging strategy has only negligible behind this concept is that a three step system indicating
impacts on user comfort, as the number of times with full the state of grid congestion is established for distribution
battery depletion during the trip increases only from 0.91 grids. In this regard, the green phase signalizes that no grid
% to 0.94 %. However, it enables the EV to be charged congestion exists and the DSO is able to keep voltage and
also during times when a grid congestion due to excess loading limits within safe boundaries by his own means,
generation exists and may therefore offer an additional such as voltage control with tap changers or reactive
income source for the EV user. Figure 3 shows a typical power control. During the red phase, on the contrary, the
charging profile during one summer week, as opposed to DSO must intervene by curtailing active power (load or
the daily normalized PV output of a 1 kWp PV plant. generation, depending on the nature of the problem) to
stay within the defined limits. Unrestricted electricity
Indicator Assumption trading is interrupted. In between the two an amber phase
Battery capacity 60 kWh is defined, where a potential bottleneck is predicted in a
Fuel economy 20 kWh/100 km defined network segment. In this phase, the DSO calls
Driving range 300 km upon flexibility that is offered by market participants, to
Charging power at home 11 kVA prevent the red phase and allow a more economical cost-
Charging efficiency 98 % allocation of flexibilities.
Power factor 0.98 var consuming
EV share of total cars 7.6 %
Number of EVs 346

Table I: Assumptions on the electric vehicle parameters

3. Description of BDEW traffic
light concept
3.1. Definition of green, amber and red phase
In two papers [10], [11] from 2015 and 2017, the German
Association for Energy and Water Industries (BDEW)
Table II: Permissible voltage and current limits for each traffic light phase
In this project, the traffic light concept has been put into
practice on the specific described distribution feeder
to quantify its effect and test its ease of use. During the
project, a real-life example of this traffic light concept is
planned to be set up. To allow for a voltage deviation of
up to 0.05 p.u. in the LV grid, an applied design criteria of

Figure 3: Normalized EV charging profile per EV in blue, normalized PV profile per kWp PV output in orange, both for a summer week

Cigre Science & Engineering • N°13 February 2019

127
the local DSO, a maximum voltage band of +/-5 % has to
be kept within the MV feeder. The green phase is defined
as being within 80 % of the permissible voltage and
overloading limits, while the red phase is evoked if the
permissible thresholds are exceeded. This sets the criteria
for the different traffic light phases according to Table II.
With a better knowledge of voltages in the MV and LV
networks, e.g. by means of a wide area voltage control,
these limits can be expanded, reducing the need for
flexibility. As can be seen in Figure 4, in the 2030 scenario
upper voltage limits are consistently violated during day
times in summer. This results also into loadings above 80
% but not exceeding 100 %. No undervoltage violations
are observed as the penetration of EVs and heat pumps is
too low to cause any problems.
Figure 4: Categorization of maximum and minimum voltage as well as loading
into green, amber and red phase of the BDEW traffic light system
for the 2030 scenario
3.2. Flexibility list
The flexibility list is one of two mechanisms described in
Cigre Science & Engineering • N°13 February 2019
128
[11] to put a flexibility market into practice. The flexibility
list describes within which boundaries, or flexibility bands,
the active power output of all flexible users (here heat
pumps, EVs, and PV) has to operate. Hereby, it specifies
the maximum feed-in (Pmin) and demand (Pmax) that
limits the combined output power of all flexibility
providers, adjusted by their respective sensitivity, to keep
within the allowed voltage and overloading (current)
limits. An example is shown in Table III.
The voltage sensitivity, shown in Figure 5, is based on
the resistance between the HV/MV substation and every
MV/LV substation. Therefore, it shows the effect on
the voltage for each flexibility option connected to the
substations. That is, if a flexibility option at the end of the
feeder (substation 29, see Figure 5) cannot be activated,
roughly triple the amount must be activated at substation
11 to gain the same effect on voltage. Only the effect on
the main branch of the feeder was assessed, allowing side
branches (substation 30 – substation 48) to be aggregated
with respect to the main branch. This is a reasonable
simplification as voltage problems do not occur on the
side branches due to the large PV plant connected to
substation 22 highly influencing the voltage. However,
long side branches with large consumption or generation
would necessitate a further split up of the flexibility list
for the different network sections.
The current sensitivity describes the impact of each
option on the thermal bottleneck in the network, i.e. the
element that receives the highest loading. The location of
the bottleneck may change depending on the feed-in and
loading situation, and even multiple bottlenecks may have
to be considered. In the analyzed network, the location
is typically between substation 02 and 03, as the cable
sections before substation 02 have larger cross-sections.
Hence, all substation except number 01 and 02 have
a current sensitivity of close to 1, deviating from unity
only due to additional line losses. These line losses can
negatively affect the line loading in the case of loads
(sensitivity > 1) or positively it in the case of generation
(sensitivity < 1).
To give an example: 1 MW of feed-in power at the end of
the feeder (substation 29) reduces the voltage flexibility
band by 1 MW (sensitivity = 1), while it reduces the
current flexibility band by only 0.890 MW as line losses
between substation 03 and 29 have to be taken into account.
 Figure 5: Determination of voltage sensitivities for each substation
Likewise, 1 MW of feed-in power at substation 03 would
result in a reduction of only 0.038 MW for the voltage
flexibility band but 1 MW for the current flexibility band.
Table III: Flexibility list for the analysed distribution grid with flexibility
band and sensitivities of flexibility providers. Pmin specifies the flexibility
band for demand, Pmax for generation.
Cigre Science & Engineering • N°13 February 2019
129
4. Flexibility control and
activation
In the considered network, only problems due to feed-
in of PV plants occur. The penetration of heat pumps
and EVs is too low to cause any problems during high
demand. Therefore, only flexibility measures to alleviate
excess generation are described. The flexibilities are
activated based on their cost as well as their sensitivity
to reduce the overvoltage or overloading problem. The
costs have been chosen according to Table IV. Costs for
PV curtailment have been obtained based on their feed-
in tariff. For heat pumps and electric vehicles no such
data is available and have been chosen arbitrarily to show
the principles of operation with a flexibility list. With
the assumed flexibility costs, an electric vehicle with a
sensitivity of 1 will be activated before a heat pump with
sensitivity of 0.3.
For each time step, flexibility measures will be activated
as long as the flexibility band (see Chapter 3.2) is violated,
starting with the cheapest flexibility measure.
 Figure 6: Flexibility activation of heat pumps, electric vehicles and PV curtailment for two days in March

Flexibility measure Price has been assumed to be about 2 and 4 times more expensive
Heat pumps 0.025 €/kWh
than heat pumps and EVs, respectively. However, PV
curtailment with a sensitivity of close to 1 will still be
Electric vehicles 0.050 €/kWh
chosen over heat pump and EV activation with a sensitivity
MV PV curtailment 0.106 €/kWh
of e.g. 0.1. Also, the provided flexibility by heat pumps
LV PV curtailment 0.122 €/kWh and EVs is limited, and curtailment can be unavoidable in
Table IV: Costs for flexibility provision some situations. However, the flexibility list ensures that
the most cost-effective options are selected.
4.1. Heat pump flexibility
A 200-litre storage tank for the heating system was 5. Simulation results
assumed for every heat pump. The temperature of the Following the price list as provided in Table IV, most heat
storage can be raised by up to 10 °C, providing a certain pump flexibility is provided first, followed by the electric
flexibility to increase the electrical consumption of the vehicle flexibility, and using PV curtailment as the least
heat pump. This temperature difference is provided by preferred option. Figure 6 shows the amount of flexibility
an increased output of the 3 kWel heat pump. If the time needed and how the different flexibility options serve this
for flexibility provision has passed, the temperature is flexibility over time. As can be seen, heat pump flexibility
gradually lowered again by the heating and hot water is activated first but has only a small overall capability
demand of the household. of taking up excess generation. Subsequently, electric
4.2. Electric vehicle flexibility vehicles are activated next. As the charging process may
take up to a few hours, the uptake of excess energy is
During normal operation EVs at home are only charged spread out over time. Additionally, new EVs are returning
if their SOC falls below 50 %. To provide flexibility
home and offer additional flexibility at later time points.
for congestion problems, also available EVs with SOCs
Nevertheless, a large part of excess generation needs to be
above 50 % can be charged. To prevent frequent charging
curtailed on the first day, as heat pump and EV flexibility
interruptions due to changes in the grid state, the charging
is not high enough. On the second exemplary day, less
process is not interrupted until the SOC has reached again
flexibility is needed. Hence, curtailment can be reduced
100 % or the car is departing.
to nearly zero.
4.3. PV flexibility (curtailment)
As a result, the voltage is reduced below its threshold
As a last option, PV generation is curtailed. Curtailment of 1.04 p.u. The deviation from the 1.04 p.u. can be

Figure 7: Voltage profile with and without flexibility provided for two days in March

Cigre Science & Engineering • N°13 February 2019

130
 Figure 8: Voltage profile with and without flexibility provided for the entire year
attributed to changes in reactive power and the change
of voltage sensitivity due to different loading and voltage
levels. Figure 8 shows that the voltage threshold is kept
throughout the whole year.
Table V shows the overall contribution of the different
flexibility options towards the overvoltage reduction for
one year. With heat pump shares of 11.9 % and EV shares
of 7.6 %, curtailment can be reduced by 15 % while costs
can be reduced by 6 % with the cost assumptions in Table
IV. Furthermore, most curtailment is provided by the large
PV plant as it is located towards the end of the feeder and
can be curtailed for a cheaper price than LV connected PV
plants. The necessary PV curtailment is below 0.04 % for
any PV plant in both cases.
6. Conclusion and discussion
The paper shows a concept how different flexibility
options can be integrated into a single framework, to
select the most cost-effective option to mitigate local grid
congestion problems (voltage violations and overloading).
The framework is called the BDEW traffic light system,
proposed by the German Association of Energy and Water
Industries (BDEW). The concept is applied to a medium
voltage feeder for the year 2030, with high PV penetration
Table V: Resulting contribution and costs of the different flexibility technologies for one year
Cigre Science & Engineering • N°13 February 2019
131
and new flexible demand, namely heat pumps and electric
vehicles. To provide flexibility, heating demand and EV
charging can be delayed as well as PV active power
output curtailed. Based on the cost of providing flexibility
and the sensitivity that the respective flexibility has on
the grid congestion problem, the least expensive options
are chosen. As a result, voltage levels are consistently
kept under the defined threshold of 1.04 p.u. Further, PV
curtailment is reduced by 15 % and costs by 6 % compared
to the case where only PV curtailment is applied.
The concept is not limited to the flexibilities integrated
here. Any other flexibilities, such as batteries or combined
heat and power units, can be added to participate in this
flexibility market. With a gradual increase of generation
and new loads, the costs for flexibility will increase, up
to the point where grid extension or other options offer
cheaper alternatives to the DSO. Hence, such upgrades
are effectively delayed or in some cases fully prevented.
To facilitate the BDEW traffic light concept, a high
degree of coordination is needed, e.g. by means of smart
meters. The flexibility list including the sensitivity matrix
can change based on the demand/generation situation and
the network topology, which needs to be continuously
communicated with the flexibility providers as well as the
contracted flexibility.
Furthermore, it should be noted that enough flexibility
liquidity needs to be available to ensure operability and
competitiveness. Lastly, the activation of flexibility should
not result in violations in the lower voltage levels. Hence,
it is foreseen that the BDEW traffic light will encompass
all voltage levels in the future.
7. Bibliography
[1] EN 50160, “Voltage characteristics of public distribution systems,”
2010.
[2] “DESIGNETZ.” [Online]. Available: https://[Accessed: 26-Mar-
2018]. [3] OpenStreetMap contributors, “Planet dump
retrieved from https://planet.osm.org.” 2018.
[4] M. Koch et al., “BWP-Branchenstudie 2015,” p. 44, 2015.
Cigre Science & Engineering • N°13 February 2019
132
[5] Joint Research Centre, “Photovoltaic geographical information
system,” 2018. [Online]. Available: . [Accessed: 26-Mar-2018].
[6] E. McKenna and M. Thomson, “High-resolution stochastic
integrated thermal-electrical domestic demand model,” Appl.
Energy, vol. 165, pp. 445–461, 2016.
[7] R. Follmer, B. Lenz, B. Jesske, and S. Quandt, “Mobilität in
Deutschland 2008,” Tempo, p. 214, 2008.
[8] B. Lunz, H. Walz, and D. U. Sauer, “Optimizing vehicle-to-grid
charging strategies using genetic algorithms under the consideration
of battery aging,” 2011 IEEE Veh. Power Propuls. Conf. VPPC 2011,
2011.
[9] G. Lacey, T. Jiang, G. Putrus, and R. Kotter, “The effect of cycling on
the state of health of the electric vehicle battery,” Proc. Univ. Power
Eng. Conf., 2013.
[10] BDEW German Association of Energy and Water Industries,
“Smart Grid Traffic Light Concept,” no. March, 2015.
[11] BDEW, “Konkretisierung des Ampelkonzepts im Verteilungsnetz,”
2017.
Insulation Characteristics in DC-GIS:
Surface charge phenomena on epoxy
spacer and metallic particle motions

T. YASUOKA*, Y. HOSHINA, Toshiba Corporation

M. SHIIKI, M. TAKEI, Toshiba Energy Systems & Solutions Corporation
A. KUMADA, K. HIDAKA, The University of Tokyo
Japan

Summary
In order to develop a DC-GIS with high insulation
reliability and compact size, it is important to clarify the
insulation characteristics of epoxy spacers with surface
charge accumulation and metallic particle motions under
a DC voltage.
A measurement system was developed for spacer
charging to investigate surface charge characteristics on
SiO2-filled epoxy in SF6 gas over 100 hours. Applied
voltage, polarity and water content (ppmv) in SF6 were
varied to identify charge accumulation phenomena. As
a result, it is suggested that the water content causes
irregular charge accumulation which cannot be simulated
with a numerical analysis taking SF6 gas carrier, electric
field dependent volume conductivity in epoxy and space
charge in epoxy into account. The tangential electric
field on a spacer that is intensified by such an irregular
charge was evaluated, concluding that the irregular
charge increased the electric field by only 20%, hardly
influencing flashover characteristics in an actual spacer
under a normal operating voltage in a DC-GIS.
The insulation performance against metallic particles
was improved by a new insulation coating for the inner
surface of the grounded (GND) enclosure and by particle
traps on a high- voltage (HV) conductor and at a GND
enclosure side. This new coating improves particle lift- off
characteristics by about 10%. the structure of the HV side
particle trap was optimized based on fundamental research
on particle motion (firefly phenomena), and a deep hand-
hole structure with a low electric field was developed for
the GND side trap. The experimental results show that a
particle falling from the HV side trap to the GND side trap
can no longer lift-off. This means the particle trap makes a
DC-GIS a highly reliable equipment.
1. Introduction
Recently, high-voltage direct current (HVDC) systems
have attracted attention for their efficient electric power
transmission. A DC-gas insulated switchgear (GIS) plays
an important role as a DC switchyard in a HVDC system,
because the DC-GIS has high insulation reliability and
compact size [1]. However, insulation phenomena under
a DC voltage have different characteristics from those
under an AC voltage (Figure 1) [2, 3, 4, 5]. A surface
charge accumulates on the epoxy spacer in SF6 gas, i.e.,
solid-gas interface, due to the electric field transition from
initial capacitive distribution to stationary distribution
with a DC voltage. This accumulated charge might cause
the reduction of breakdown (BD) characteristics in not
only a DC voltage but also a superimposed lightning
impulse voltage on a DC voltage. A wire metallic particle
also strongly affects insulation characteristics. The firefly
phenomenon, whereby a particle hovers around a high-
voltage conductor after lifting-off from the enclosure, is
well known under a negative DC voltage [5, 6, 7]. BD
voltage decreases significantly, when the lighting surge
enters under a firefly situation. Thus, to develop a DC-
GIS, it is important to clarify the insulation characteristics
of surface charge phenomena on epoxy spacers and
metallic particle motions under a DC voltage.

* takanori.yasuoka@toshiba.co.jp

KEYWORDS
HVDC - DC-GIS - Surface Charge - Epoxy - SF6 Gas - Metallic Particle - Particle Lift-off - Coating - Firefly - Particle
Trap

Cigre Science & Engineering • N°13 February 2019

133
 Figure 1: Insulation characteristics of spacer charging and metallic particles in GIS under AC and DC
2. Surface charge on an epoxy
spacer under a dc voltage
The surface charge on an epoxy spacer is an important
issue in a SF6/epoxy composite insulation system under a
DC voltage, because the surface charge might decrease the
BD voltage. To identify its characteristics, spacer surface
charging was investigated. We developed a measurement
system to evaluate surface charge density on a downsized
SiO2-filled epoxy spacer with a diameter of 90 mm in SF6
0.5 MPa-abs under a DC voltage (Figure 2) [8]. The DC
voltage application time was over 100 hours.
A DC voltage was applied to a HV electrode with GND
electrodes in a closed configuration. After the DC voltage
was turned off, the GND electrodes opened, and the
surface potential induced by the accumulated charge
on the spacer was measured with electro-static probes
connected to a high-input impedance electrometer.
The number of measuring points was 3444 for both a
sloping surface and a flat surface on the model spacer.
Generally, the induced surface potentials we measured
during the scanning process were influenced by all of
the accumulated charge on the spacer. Therefore, the
inverse calculation method, including signal processing,
was developed to obtain high spatial resolution of surface
charge density. This method is described in detail in
previous papers [8, 9]. Surface charge characteristics are
very sensitive to surface condition. Before an experiment,
the model spacer is conditioned by drying it for 24 hours
Figure 2: Surface charge measurement systems: a) Experimental setup; b) Measurement parts; c) Model spacer (SiO2-filled epoxy)
Cigre Science & Engineering • N°13 February 2019
134
at a temperature of 80ºC. The water content in SF6 gas
was monitored by dew point measurement system during
DC voltage applications.
To identify the mechanism of charge accumulation, applied
voltage and polarity were varied. Moisture content (ppmv)
in SF6 gas was also changed by controlling the absorbent
or injecting water into the chamber, etc. The water content
was affected by ambient temperature, so values for water
content (ppmv) are indicated with a range of values, e.g.
30 - 80 ppmv. Surface charge density was calculated with
a simulation model taking non-linear carrier behaviour
in SF6 gas (ion drift) and electric field dependent volume
conductivity in epoxy and space charge in epoxy into
account for comparison with measured results and to
identify the mechanism. Finally, the risk of an accumulated
charge against the BD characteristics was evaluated.
2.1. Surface charge characteristics on SiO2-filled epoxy
Figure 3 shows time variations of surface charge
accumulation on the sloping surface (Figure 3a) and
the flat surface (Figure 3b) as an example. Under these
conditions, the accumulated charge was conspicuous after
applying a DC voltage for 70 hours. The homo-charge was
observed around the HV electrode and the hetero-charge
was observed around the GND electrode especially on
the sloping surface. The difference between the charging
patterns on sloping surface and the flat surface depended
on electric field distribution and surface conditions. This
tendency is discussed in Section 2.2.
 Figure 3: Time variations of surface charge accumulation; V = +60 kV; water content 30 - 80 ppmv; SF6 0.5 MPa-abs: a) Sloping surface; b) Flat surface

The dependence of applied voltage on charge accumulation
was investigated by changing the DC voltage from +30 kV
to +60 kV (Figure 4a and 4b). As expected, the value of
the surface charge density was increased, when the applied
voltage was increased. The irregular charging pattern with
positive and negative pair was seen in the direction of 3
o’clock around the middle area on the sloping surface.
This might have been caused by tiny metallic particles or
a non-uniformity in the surface conditions.
The polarity effects on charge accumulation were
investigated at +60 kV and -60 kV (Figure 5a and 5b).
Basically, even if the polarity of the DC voltage is
changed, the surface charge accumulated at the same
location with a reverse polarity. However, in the case of
a negative polarity, charge density of the homo-charge
around the HV conductor has a larger value than in the
case of a positive polarity.
The reason is that a micro-discharge or a field emission
from the cathode with a negative polarity might contribute
to a larger accumulated charge.
The influences of water content in SF6 gas were investigated
by changing the water contents from 4, 30 – 80, 200 –
300, 500 – 650 ppmv. The surface charge density on the
sloping surface after about 120 hours is shown in Figure
6a. Time variations of charge density depending on water
content are indicated in Figure 6b. From these figures, the
more the water contents increases, the more the irregular
surface charge, not the concentric circle profile, near the
HV electrode increases. Water in SF6 gas and adhering to
the epoxy surface or electrode acts as a carrier in a high
electric field [10], influencing the charge accumulation on
the epoxy. In these experimental conditions, the saturation
time of charge accumulation mainly influenced by water
is estimated to be about 200 hours.

Figure 4: Surface charge accumulation for different applied voltage cases; water content 200 - 300 ppmv; SF6 0.5 MPa-abs: a) +30 kV; b) +60 kV

Figure 5: Polarity effects of surface charge accumulation; water content 200 - 300 ppmv; SF6 0.5 MPa-abs: a) +60 kV; b) -60 kV

Cigre Science & Engineering • N°13 February 2019

135
 Figure 6: Influence of water contents in SF6 gas on surface charge accumulation; +60 kV, sloping surface; SF6 0.5 MPa-abs:
a) surface charge density after about 120 hours;
b) Time variation of charge density, the value on the vertical axis is the average value from the top 10 surface charge densities.

2.2. Simulation of surface charge accumulation volume conductivity which is estimated by polarization
and depolarization current measurement over 1 month.
A simulation was performed for the comparison with
A homo-charge near the cathode and the anode was
measured surface charge density and identification of
observed with the pulsed electroacoustic analysis (PEA)
surface charge mechanism. In this simulation (Figure 7),
measurement, and it was found that the measured space
case 1 and case 2 are applied to calculate surface charge
charge almost corresponds to the integrated polarization
on the epoxy in a 2-D axi-symmetric geometry. In case 1
current measurement (vs time, over 1 month) [12]. In
(improved model), SF6 gas is not a resistor but a carrier
addition, the current conduction characteristic of the the
source, i.e. positive and negative ion densities, which
SF6 gas/epoxy interface was determined by the appropriate
are the origin of current drifted by the electric field, are
value based on surface resistivity measurement. This
calculated based on the convection-diffusion equation
simulation method is described in detail in a previous
including the terms of ion-pair generation and ion-ion
paper [13]. In case 2 (simple model), dielectric constant
recombination.
ε and volume conductivity σ in SF6 gas and epoxy are
(1) constant values. These cases are applied to a simulation
implemented in the COMSOL Multiphysics ®.

(2) The simulation results are shown in Figure 8, positive

(3)
Ion-pair generation rate is assumed to be No = 40 IP/
(cm3·s), ion-ion recombination kr = 3.4×107 cm3/s, ion
mobility µ+ = µ- = 0.08 cm2/(V·s), diffusion coefficient
D+ and D- are determined using Einstein’s equation
[11]. SiO2-filled epoxy has an electric field-dependent
and negative ion densities in SF6 gas after applying a
DC voltage for 120 hours (Figure 8a) and surface charge
density on the sloping surface for different cases (Figure
8b) as an example. Big difference is found in the charging
profile and time to reach the DC steady state condition
between case 1 and case 2, i.e., amount of accumulated
charge was smaller in case 1 than in case 2, and time to
reach a steady state was longer in case 1 than in case2.

Figure 7: Model and physical properties for simulating surface charge density in case 1 and case 2

Cigre Science & Engineering • N°13 February 2019

136
 Figure 8: a) Positive and negative ion density after +60 kV 120 hours in SF6 0.5 MPa-abs; b) Simulated surface charge density on the sloping surface, case
1(+60 kV, including carrier source in SF6 electric field dependence of σV in epoxy and space charge in epoxy), case 2(+60 kV, εr = 4 and σV = 10-16 [S/m]
in epoxy, εr = 1 and σV = 10-19 [S/m] in SF6 gas)

A comparison of surface charge between measurement
and simulation was performed as shown in Figure 9.
From this figure, even in case 1, which is the improved
model, the measurement result could not be simulated. It
is speculated that a micro-discharge at the triple junction
on the HV conductor might cause positive charge, even
when the electric field is attenuated by a cast-in electrode
in epoxy and a shield electrode near the spacer. An
irregular charge near the GND electrode, not a concentric
circle profile, might be caused by micro-discharge at the
edge with a locally intensified electric field on the GND
electrode. The non-uniformity of the surface conditions
on the epoxy surface might lead to a spot-type surface
charge at the middle area on the epoxy. These phenomena
are emphasised by the water content in SF6 and applied
electric field as described above. To achieve a more
reasonable simulation, the influences of water content as
a carrier and conduction at the interface on SF6 gas/epoxy
should be taken into account.
2.3. Evaluation of intensified electric field caused by
surface charge accumulation
Basically, a surface charge accumulation on the epoxy
cannot be avoided under a DC. So, it is very important to
estimate the risk of an accumulated surface charge against
the BD characteristics. Risk was evaluated by calculating
tangential electric field strength on the epoxy spacer (Figure
10). Tangential electric field distribution without a surface
charge, initial capacitive distribution, is shown in Figure
10a, with surface charge, the worst case in our experiment,
shown in Figure 10b. Through an evaluation based on an
electric field calculation, it was found that the irregular
charge intensified the tangential electric field by about
20% compared to the initial capacitive situation without
a surface charge. But even such intensified tangential
electric field was significantly lower than that of a surface
flashover field [14]. It is concluded that the irregular charge
hardly influences flashover characteristics in an actual
epoxy spacer under a normal operating DC voltage and a
superimposed lightning surge on DC voltage in a DC-GIS.
3. Metallic particle motions in sf6
under a dc voltage
A wire metallic particle strongly affects insulation
characteristics in a DC-GIS. In the case of a negative DC
voltage, the firefly phenomenon, whereby a particle hovers
around at a high-voltage conductor, is well known [5, 6,
7]. BD voltage decreases significantly, when a lightning
surge is applied under a firefly situation. It is necessary
to measure this risk and make the particle harmless. The
policy for improving insulation performance against

Figure 9: Comparison between measurement and simulation; +60 kV 120 hours in SF6 0.5 MPa-abs: a) Measured surface charge density on the sloping
surface; b) Simulated surface charge density on the sloping surface in case 1(+60 kV, including carrier source in SF6, Electric field dependence of σV in
epoxy and space charge in epoxy) and case 2(+60 kV, εr = 4 and σV = 10-16 [S/m] in epoxy, εr = 1 and σV = 10-19 [S/m] in SF6 gas)

Cigre Science & Engineering • N°13 February 2019

137
 Figure 10: Impact of surface charge accumulation on a tangential electric field at the sloping surface; +60 kV
in SF6 0.5 MPa-abs: a) without surface charge = capacitive distribution; b) with surface charge; after 120 hours;
water content 500 – 650 ppmv

metallic particles in DC-GIS is shown in Figure 11. In the
first process, it is important to suppress particle lift-off.
A new insulation coating material for the inner enclosure
was developed to improve particle lift-off characteristics.
In the second process, trapping particles is also important
after they lift-off. A particle trap is equipped at the HV
and GND side for making a particle fall into the GND
side trap just after trap on HV side. Thus, the worst-case
situation that wire metallic particles adhere to the epoxy
insulator can be impeded.
3.1. New coating material for inner surface of
enclosure
Three mechanisms cause particles to lift-off from the
enclosure under a DC [5, 15]; first one is charging-up the
particle from the coating or the SF6 gas due to polarization,
second is charging-up the particle by partial discharge
in triple junction with high electric field strength, third
is an acceleration to the particle, such as operation of a
disconnecting switch, etc. To improve particle lift-off
characteristics, coating materials should have appropriate
values for dielectric constant, volume conductivity,
hardness, surface roughness, etc. [16]. Figure 12 shows
the experimental setup for investigating particle lift-
off. Aluminium particles 3 mm long and 0.25 mm in
diameter were set on the bottom of a mock enclosure in
SF6 0.45MPa-abs. The acceleration of the particle was
generated by the vibration of a hammer, and this value was
measured with a sensor. Positive and negative DC voltage
were applied. The specification of the dielectric coating is
shown in Table 1. A ZnO semi-conductor is added to the
epoxy-based coating to control physical properties.

Figure 11: Policy for improving insulation performance against metallic particles in DC-GIS

Figure 12: Experimental setup for investigating particle lift-off from the enclosure: a) Sketch of Sectional view; b) Image from observation window

Cigre Science & Engineering • N°13 February 2019

138
 Table 1: Specifications of dielectric coatings for inner surface of enclosure in DC-GIS

Materials Thickness of coating Dielectric constant Volume conductivity

4×10-14 [S/m]
Conventional Epoxy based coating 50-100 [µm] 5.2 @100Hz, RT
@ 0.1 kV/mm, RT
5×10-14 [S/m]
New Epoxy + ZnO semi-conductor 5wt% 50-100 [µm] 8.5 @100Hz, RT
@ 0.1 kV/mm, RT

Table 2 shows the results for particle lift-off
characteristics without vibration in conventional
and new dielectric coatings. Particle lift-off was not
observed under these experimental conditions with
applied DC voltage 1.0 p.u: 4 hours, 1.5 p.u.: 2 hours,
1.75 p.u.: 1 minute, under positive and negative DC. A
contact point between particle and coating, conductivity
of coating and effective conductivity of SF6 gas play
important roles in particle lift-off without vibrations,
because these characteristics dominate the charging-up
of the particle under normal operating and dielectric
test condition in DC-GIS where the electric field lower
than partial discharge inception field is formed at the
triple junction. Based on an investigation of contact area
measurement, the contact area is extremely small, thus,
the charge from the coating via contact and the SF6 gas
into the particle is likely to be sufficient to prevent the
particle from lifting off. It might be effective even after
Table 2: Results of particle lift-off characteristics for different coatings without vibrations
the coating polarizes to equilibrium, i.e., DC steady
state condition.
Particle lift-off characteristics with vibrations were also
investigated. Figure 13 shows the relationship between
acceleration and lift-off probability for different coating
materials at positive (Figure 13a) and negative applied DC
voltage (Figure 13b). From these figures, it was found,
as follows, that particles lifted off when the acceleration
exceeded 20 G, particles in a negative applied voltage
lifted off more easily than those of in a positive applied
voltage, and particles with a higher DC voltage of 1.0 p.u.
lifted off more easily than those of 0.9 p.u.. As a result,
the particle lift-off probability on the new coating at 1.0
p.u. is a same value as on a conventional coating at 0.9
p.u., meaning that the particle lift-off electric field was
improved by about 10%. Consequentry, the DC-GIS was
downsized with the new coating.

Polarity results

Positive No liftoff in 1.0 p.u: 4hours, 1.5 p.u.: 2hours and 1.75 p.u.: 1minute
Conventional coating
Negative No liftoff in 1.0 p.u: 4hours, 1.5 p.u.: 2hours and 1.75 p.u.: 1minute

Positive No liftoff in 1.0 p.u: 4hours, 1.5 p.u.: 2hours and 1.75 p.u.: 1minute
New coating
Negative No liftoff in 1.0 p.u: 4hours, 1.5 p.u.: 2hours and 1.75 p.u.: 1minute

Figure 13: Particle liftoff characteristics for different coatings with vibrations; relationship between acceleration and liftoff probability:
a) Positive polarity; b) Negative polarity

Cigre Science & Engineering • N°13 February 2019

139
 Figure 14: a) Method of evaluating firefly phenomena: b) Firefly motion; combined image observed with a high-speed video camera

3.2. Development of particle trap
The particle trap, as seen in Figure 11, is equipped in
order to suppress such a situation in which wire metallic
particles adhere to the epoxy insulator. There are two types
of particle trap: HV side trap and GND-enclosure side
trap. The former is placed on the HV conductor adjacent
to a spacer and the latter is located on the GND enclosure
just underneath the former one. To develop an effective
particle trap, optimization of the structure of the HV side
trap takes firefly phenomena into account, so, fundamental
research was performed to investigate firefly phenomena
by using actual size bus-bar model. Figure 14a shows the
method for evaluating firefly characteristics; duration and
moving distance were investigated by changing applied
DC voltage and length of wire metallic particle.
The combined image observed by a high-speed video
camera in Figure 14b provided the following particle
motion during firefly phenomenon. The particle hovers
around randomly mainly in the longitudinal direction
(sometimes in the circumferential direction) near the HV
conductor with a visible partial discharge between the
HV conductors and has a precession. It was identified
that the key driver maintaining the firefly motion was
a continuous corona wind force at the particle tip of the
enclosure side. Based on this result, the reason that firefly
was not seen with a positive DC can be explained; the
corona wind derived from the partial discharge have to be
occurred just after first hitting the conductor from lift-off
to maintain firefly state. On the other hand, the particle
charge changes to the opposite polarity after hitting the
conductor from lift-off, i.e., particle wants to return to
the enclosure. In negative polarity, a statistical time lag
of partial discharge is very short, because field emissions
from particle tip are a source of initial electrons; however,
in a positive polarity, the statistical time lag is long [17],
i.e., particles cannot be in a firefly state and return to the
enclosure. In addition, the electric field on the particle tip
becomes lower at a position away from HV conductor, so,
it is very difficult to return to the HV conductor by corona
wind force caused by the partial discharge. The process
whereby firefly phenomenon ends was also identified
as follows: the axis of precession becomes unstable
(increasing the angle) due to the unbalance between
corona wind force and rotational torque, meaning the
electric field on the particle tip decreases, so particles
cannot maintain a firefly state.
The firefly duration depends strongly on particle length
and applied DC voltage as can be seen in Figure 15a.
It was also found that the moving distance of a firefly
particle has a positive correlation with duration, as shown
although this tendency has a large dispersion, in Figure
15b. These results contributed to optimizing the structure
of the HV side particle trap.

Figure 15: Firefly characteristics on an actual size gas insulated bus-bar: a) Characteristics of firefly duration: b) Relationship between firefly duration and
moving distance; *1 when firefly duration was short, the PD signal was not detected. In that case, duration was decided as a value of the minimum time
resolution of the high-speed camera. *2 Duration was determined with the PD signal. *3 when the particle was out of the viewing angle of the high-speed
camera, the moving distance was determined by confirmating particle position after opening the enclosure.

Cigre Science & Engineering • N°13 February 2019

140
 Figure 16: a) Experimental setup for investigating the effects of the particle trap in a DC-GIS bus-bar model; b) Sketch of sectional view of a);
c) Metallic particle, particle trap at HV side and GND side by observing with a high-speed video camera (still image)

The effectiveness of the developed particle trap was
investigated by actual Bus-bar model of a DC- GIS.
Figure 16 shows the experimental setup. The structure of
the high-voltage side trap was optimized so that a fire-
flying particle easily falls under normal operating voltage.
The grounded- enclosure side trap has a deep hand-hole
structure and low electric field is formed around it. Thus,
a particle falling from the high-voltage side trap to the
grounded-enclosure side trap can no longer lift off.
The particle positions before and after the experiment
performed in the particle trap investigations are shown in
Figure 17. In this experiment, strong accelerations were
applied to the particle by vibrations in order to have the
particle lift off forcibly. It was found that the number
of particles adhering to the epoxy spacer was zero, and
many particles were trapped at the GND side. There were
particles that were not trapped at the GND side; however,
it is assumed that particles outside the trap are also trapped
at the GND side eventually. This means that the particle
trap makes the DC-GIS a highly reliable equipment.

Figure 17: Particle positions before and after experiment performed for the particle trap investigations; particle φ = 0.25 mm,
length 3 mm; positive and negative DC voltage

Cigre Science & Engineering • N°13 February 2019

141
4. Conclusion
In order to develop a DC-GIS with high insulation
reliability and a compact size for switchyards of HDVC
transmission systems, surface charge characteristics
and metallic particle motions under a DC voltage were
investigated. These data can be applied to insulation
designs for a new 250 kV DC-GIS in Japan [1].
5. Bibliography
[1] M. Matsubara, Y. Abe, T. Yasuoka, M. Shiiki, M. Takei and T.
Karube, “ The Verification of Withstand Voltage Performance for
250 kV DC GIS” (Annual Conference of Power and Energy Society,
IEE of Japan, Tokyo, Japan, 2017, 324)
[2] R. Gremaud, F. Molitor, C. Doiron, T. Christen, U. Riechert, U.
Straumann, B. Kallstrand and K. Johansson, “Solid Insulation in
DC Gas-Insulated Systems” (Cigre Session 45, Paris, France, 2014,
p. D1_103_2014)
[3] A. Winter, J. Kindersberger, M. Tenzer, V. Hinrichsen, L. Zavattoni,
O. Lesaint, M. Muhr and D. Imamovic, “ Solid/Gaseous Insulation
Systems for Compact HVDC Solutions” (Cigre Session 45, Paris,
France, 2014, p. D1_102_2014)
[4] Cong-Thanh VU, L. Zavattoni, P. Vinson and A. Girodet, “DC-
GIS: Importance of simulations for the design and testing” (Cigre
Session 46, Paris, France, 2016, p. D1_102_2016)
[5] CIGRE Task Force D1.03.11, “Gas insulated systems for HVDC:
DC stress at DC and AC systems” (Cigre Technical Brochure, No.
506, 2012)
[6] S.A. Boggs, K. Nojima and T. Yasuoka “Computation of Horizontal
Force Normal to the Axis of a Spheroidal Conducting Particle in
a Field Gradient” (IEEE Trans. Dielect. Electr. Insul., Vol. 23, pp.
1649-1651, 2016)
[7] S.A. Boggs, K. Nojima and T. Yasuoka “Computation of
Particle Motion in Gas-Filled Systems” (Proceedings of the IEEE
International Symposium on Dielectrics, Montpellier, France,
2016, pp. 418-422)
[8] H. Iwabuchi, S. Matsuoka, A. Kumada, K. Hidaka, Y. Hoshina,
T. Yasuoka and M. Takei, “Influence of Tiny Metallic Particles on
Charge Accumulation Phenomena of GIS Model Spacer in High-
Cigre Science & Engineering • N°13 February 2019
142
pressure SF6 Gas” (IEEE Trans. Dielect. Electr. Insul., Vol. 20, pp.
1895-1901, 2013)
[9] A. Kumada, and S. Okabe, “Charge Distribution Measurement on
a Truncated Cone Spacer under dc Voltage” (IEEE Trans. Dielect.
Electr. Insul., Vol. 11, pp. 929-938, 2004)
[10] R. Hanna, O. Lesaint and L. Zavattoni, : ”Dark Current
Measurements in Humid SF6 at High Uniform Electric Field”
(IEEE Conf. on Elec. Insul. and Diel. Phen., CEIDP, Tronto,
Canada, 2016)
[11] T. Yasuoka, H. Kumai, M. Takei, Y. Hoshina, T. Maruo, S.
Matsuoka, A. Kumada and K. Hidaka, “Numerical Simulation
of Electrical Conduction Phenomena in Various Insulating Gases
(SF6, Dry Air, CO2 and N2) “ (IEE Japan annual meeting, 6-279,
Ehime, Japan, 2014)
[12] C. Cooke, Y. Hoshina, M. Miyashita and Y. Shimizu, “Polarization
and Charge Transport in Unfilled Epoxy “ (International
Symposium on Electrical Insulating Materials (ISEIM2005), pp.
69-72 Vol. 1, Kitakyushu, Japan, 2005)
[13] R. Suda, S. Matsuoka, A. Kumada, K. Hidaka, T. Yasuoka,
H. Kumai, Y. Hoshina and M. Takei, “Charge Accumulation
Phenomena of Silica-filled Epoxy Resin Spacer under DC field”
(The 19th International Symposium on High Voltage Engineering,
Pilsen, Czech Republic,318, 2015)
[14] S. Kaneko, S. Okabe, T. Kobayashi, K. Nojima, M. Takei and T.
Miyamoto, “Evaluation of Withstand Performance for Epoxy
Insulator under Residual DC Charge in GIS” (IEEJ Transactions
on Power and Energy, Vol. 128, No. 5 P 728-734, 2008)
[15] S.A. Boggs, K. Nojima and T. Yasuoka “Electric Field-Induced
Particle Lift-Off from a Dielectric Coating into a Gas” (IEEE
Trans. Dielect. Electr. Insul., (to be submitted))
[16] T. Yasuoka, K. Nojima, M. Matsubara, M. Shiiki, M. Takei and S.A.
Boggs, “Identification of Particle Parameters through Matching
of Computed and Measured Trajectories” (2017 Annual Report
of the IEEE Conference on Electrical Insulation and Dielectric
Phenomena, Fort Worth, TX, USA, 2017)
[17] N. Wiegart, L. Niemeyer, F. Pinnekamp, W. Boeck, J. Kindersberger,
R. Morrow, W. Zaengl, M. Zwicky, I. Gallimberti, and S. A.
Boggs: “Inhomogeneous Field Breakdown in GIS-The Prediction
of Breakdown Probabilities and Voltage Part II: Ion Density and
Statistical Time Lag” (IEEE Trans. Power Delivery, Vol. 3, No.3, pp.
931-938, 1998)
Benefit and Resolution of operational
issues for information and communication
systems using virtualization techniques
in the electric power industry

H. DOI*, Central Research Institute of Electric Power Industry (CRIEPI)

D. KARIU, Kyushu Electric Power Co., Inc.
K. SAKAMOTO, Electric Power Development Co., Inc.
M. SUSUKITA, Kansai Electric Power Co., Inc.
T. SHIMA, Humming Heads. Inc.
H. NAGASAKI, JP Business Service Corporation
Japan

Summary system makes a virtualized client environment image on

Japanese electric power companies make efforts to
streamline workflow and reduce costs for maintenance
and operations as well as create secure environments
by network and information system virtualization.
This paper focuses on network virtualization and office
terminal virtualization. The first topic is concerned with
the Software Define Network (SDN) that plays a key role
in the network virtualization system. Since electric power
companies need to reduce operational and maintenance
costs, the authors examine the benefits of SDN for
network operation tasks in electric power companies and
show what network operation tasks are omitted or made
easier, because SDN realizes to automate configuration
of network devices, instead of manually configuring
ones. The second topic is a partial thin client system
based on Virtual Desktop Infrastructure (VDI). Japanese
electric power companies have introduced many business
applications, and the operation costs of virtualization
systems tend to be expensive in the case of all applications
and storage disks virtualized. Therefore, the authors
developed a new method that employs only the partially
virtualized local disks of corporate computers. This
method does not store user data for security enhancement,
so users can execute office applications on a corporate
computer. Also, the authors developed the virtualization
system for the highest necessity business application. This
a server to execute the business application and displays
it on a client browser when a client user needs to use
the application. The physical server which executes the
application for a client is not needed and server cost is
reduced. The business application can be maintained
uniformly such as version up or applying a patch for
security. Thus, security management cost is reduced, too.
The authors realized the partially thin client environments
with the system virtualization by not only reducing the
operation costs but also by securing environments.
1. Introduction
In Japanese electric power companies today, there are
efforts to make workflow more efficient, reduce
maintenance and operational costs, and create secure and
seamless access environments. As one of the methods
to realize workflow and cost efficiency, network and
information system virtualization is examined and
implemented. The Software Defined Network (SDN) [1]
is one of the major network virtualization systems. The
SDN has already been implemented in many Japanese
companies and brings benefits, for example reducing
maintenance and operation costs [2]. Also, it is expected
to obtain the same kind of benefits in Japanese electric
power companies. Thus, we examine the benefits of SDN
for network operations in Japanese electric companies.
At first, we show the network operation flows and work

* doi@criepi.denken.or.jp

KEYWORDS
SDN - OpenFlow - Virtual Desktop Infrastructure - Thin Client - Web - Operation Cost

Cigre Science & Engineering • N°13 February 2019

143
 Fig. 1 Concept of SDN’s VTN
items for the setup of new communication lines in the
Japanese electric companies. Then we study the effects
of network operation flows and items in the case of
applying SDN and show what configuration tasks can be
omitted or made easier to operate.
Next, we show our developed thin client environments
as information system virtualization. There is Virtual
Desktop Infrastructure (VDI), which is one of the
typical information system virtualization techniques
[3]. But if all business applications are virtualized on
VDI, both maintenance and operation costs become
very expensive. To ease those issues, we newly design
and develop thin client environments based on VDI by
virtualizing only the partially virtualized local disks
of corporate computers and internet viewing systems,
which is a critical business application. In this report, we
show details of our thin client environments and those
benefits. In addition, we mention the issues of these
environments which are discovered after implementing
our own environments.
2. Reduction effects of network
operation process using network
virtualization techniques in
electric power industry
2.1 SDN Overview
The SDN entails the decoupling of the control plane
from the data forwarding plane and offloads its control
functions to a centralized controller. The control plane
is comprised of GUI applications for virtual network
operations and controllers which control physical
network devices. On the other hand, a data plane is
comprised of OpenFlow available switches that process
data packet and firewalls.
OpenFlow is a standard control protocol between a
controller and a switch [4]. A controller can get switch
Cigre Science & Engineering • N°13 February 2019
144
status information or flow statistics information by the
OpenFlow protocol. Network operators can build a
virtual network on GUI application. Network devices
on the physical network are automatically configured
for the virtual network by the SDN controller. Therefore
operators can design and deploy any desired network
without knowing the physical network topology or
bandwidth restrictions.
2.2 SDN system structure example on electric power
company’s control network
There is a workflow which is a setup of new
communications lines as electric power company’s work
flows with high frequency (100 lines/day), because the
electrical equipment maintenance work. Also, there is the
Virtual Tenant Network (VTN) [5] of a SDN controller
(OpenDaylight) [6] function to realize new network
automatic setup. These VTN are expected to ease this
workflow of new lines setup in electric power company.
We suppose that SDN with VTN is implemented on an
electric power company’s network and so we show this
network structure of SDN in Fig. 1.
VTN provides virtualized conventional network
functions, such as L2/L3 switch, a gateway and so on.
Once the conventional L2/L3 network is designed on
VTN, it will automatically be mapped onto an underlying
physical network, and then configured on the individual
switch leveraging SDN control protocol. The electric
power company’s control network has many network
devices. Therefore, the number of network devices which
are needed to be configured manually becomes large and
the benefits of automatic configuration becomes large
as well.
2.3 Benefits of implementing SDN for new lines
setup work flow
We focus on workflows for the setup of new
communications lines in electric power companies to
make the benefits of SDN clear. Because the frequency
that those workflows are held is very high and it is very Table 1. Operation items for planning
important to reduce those workflows. Here are two Operation Conventional
New workflow on SDN
typical cases of new lines setup workflows. items workflow
Case 1
[case 1] To create a new network adding new physical · Confirmation of
devices existing network
[case 2] To create a new network on existing physical Device plan devices · Same
· Selection of new
devices network devices
The workflows are divided into the following 3 steps. · Confirmation of
existing line routes.
Step1: Planning Line plan · Making the · Same
specification of new
Step2: Making specifications and preparations lines.
Step3: Network device setup and execution on site Case 2

Step1: Planning · None (Controller

· Confirmation of
automatically confirms
According to a demand for new communication lines, Device plan existing network
existing network
devices
network planners start to examine the plan for new devices)
communication lines and needed network devices. · Confirmation of existing
We show the summary of operation items of step 1: · Making the
line routes.
Line plan specification of
planning for both the conventional workflow and the · Making the specification
new lines.
of new lines.
new workflow on SDN in table 1.

Table 2. Operation items for making specifications and preparations

Operation items Conventional workflow New workflow on SDN

Case 1
Making specificati on of · Making specification of new devices and VLAN or · Making specification of new devices and
devices MPLS configurations OpenFlow configurations

Making specification of lines · Making specification of new lines · Design virtual network

Order · Order new devices and making configuration · Order new OpenFlow available switches

· None (Controller automatically manages

Database entry · Database entry for new devices and lines
switches and lines)
Coordinate the setup
· Arrangement of operation time and items · Same
schedule

Registration · Registration of worker’s operations in a machine room · Same

Case 2
Making specificati on of lines · Making specification of new lines · Design virtual network

· None (Controller automatically manages

Database entry · Database entry for new lines
switches and lines)

Coordinate the setup

· Arrangement of worker’s operation time and items · None
schedule

Registration · Registration of worker’s operations in a machine room · None

Cigre Science & Engineering • N°13 February 2019

145
 Table 3. Operation items for network device setup and execution on site
Operation items Conventional workflow
Case 1
Procedure to start on site · Procedure to start the work’s operation on site
Reroute existing lines · Reroute existing lines for restarting devices
Device setup · Install and setup devices
Confirmation of configurations · Enter and confirmation configurations for new lines · Confirmation on GUI
Confirmation of device running · Confirmation of configured device running
Procedure to end · Procedure to end the work’s operation on site
Report of operation · Report and record the worker’s operation
Case 2
Procedure to start on site · Procedure to start the work’s operation on site
Reroute existing lines · Reroute existing lines for restarting devices
Confirmation of configurations · Enter and confirmation configurations for new lines
Confirmation of device
· Confirmation of configured device running
running
Procedure to end · Procedure to end the work’s operation
Report of operation · Report and record of the worker’s operation
[case 1] Network planners confirm the existing network
devices, examine new network devices and decide the
specifications and routes for each new communication
line. The conventional workflow and new workflow on
SDN are the same in this step.
[case 2] In the conventional workflow, network planners
confirm the existing network devices and the structure of
existing communication lines to decide the specifications
and routes for each new communication line. In the new
workflow on SDN, demanded communication lines or
networks are automatically mapped into an existing
network. Therefore network planners need not confirm
the structure of an existing network.
Step2: Decide on specifications and preparations If the
planned specifications and routes for new communication
lines are approved, network planners make detailed
network device specifications including configuration.
Cigre Science & Engineering • N°13 February 2019
146
New workflow on SDN
· Same
· None (Controller automatically manages lines)
· Same
· Confirmation on GUI
· Same
· Same
· None
· None (Controller automatically manages switches and
lines)
· None (Controller automatically manages switches and
lines)
· Confirmation on GUI
· None
· None
When specifications are fixed, network planners order
the network devices and arrange the workers for network
devices setup on site. We show the summary of operation
items of step 2: making specifications and preparations
for both workflows on SDN in table 2.
[case 1] In the conventional workflow, network planners
write not only about the performance of the new network
device but also configurations in the specification, such
as VLAN or MPLS configuration and so on. When the
detailed specifications are fixed, new network devices are
ordered and added to the device management database.
Also, setup schedule is coordinated and registered to
schedule management system. In the new workflow
on SDN, the main configuration items of OpenFlow
available switch is only port enable for OpenFlow and
controller IP address. These setup configurations are
easier than VLAN or MPLS ones. Network planners need
 Fig. 2 Workflows to create a new network
to make the virtual networks on VTN. A SDN controller
has the database for a network topology and devices on a
network, thus, an external device management database
is not needed.
[case 2] In the conventional workflow, network planners
write network devices configurations in the specification
and it is needed to arrange the workers for network
devices setup in a machine room. In the new workflow
on SDN, it is not needed to arrange the workers, because
the OpenFlow’s available switch need not stop for the
modification of configuration. The network planners
need to operate the virtual networks on VTN and
coordination of its operation schedule is needed.
Step3: Device setup
According to specifications, network devices are set up
and confirmed running network functions correctly in
this step. We show the summary of operation items of
step3: network device setup and execution on site for
both workflows in table 3.
[case 1] In both workflows, the workers setup hardware
devices and enter the configuration in a machine room of
the electric power office. In the conventional workflow,
workers enter the configuration for VLAN or MPLS and
confirm the running network is function correctly. In the
new workflow on SDN, workers enter the configuration
only for OpenFlow port enable and the controller IP
address and they confirm to connect between a switch and
a controller. This process is easier than the conventional
one. When a switch can connect to a controller, the VTN
function of the controller automatically sets up switches.
[case 2] In the conventional workflow, a device setup in
a machine room of the electric power office is needed to
change the main route of the existing communications
lines to a backup one. Because the workers stop
communication lines connected to a configured
Cigre Science & Engineering • N°13 February 2019
147
network device during changing the network device
configurations and restarting a configured network
device, and the workers need to check that all network
devices execute normally. But in the new workflow on
SDN, this worker’s process is omitted and the SDN
controller automatically configures the underlying
switches.
Finally, we summarize workflows to create a new network
and show reduced operation items and omitted operation
items in Fig. 2. We can see that many operation items by
workers are reduced or omitted by the SDN techniques.
As a result, network operation cost and time is reduced
when SDN is implemented on communication networks
of electric power company.
2.4 Future works
There are requirements for an electric power company’s
control network. The requirements are 3 items: delay
time, grace time for disruption and reliability. In
the conventional operation, the line bandwidth has
redundancy to avoid the network congestion and the
increase delay time. To satisfy grace time for disruption,
communication lines have backups. Reliability of
communication lines is manually calculated and checked
in the planning. In the case of SDN, requirements can be
satisfied by the same methods. But we study the new
method that requirements are automatically satisfied
by new functions of a SDN controller when demanded
communication lines are built on VTN.
3. Solving issues of vdi-based
thin client systems
Kyushu Electric and J-Power have been preparing to
implement thin client system virtualization using VDI to
improve work efficiency application and device security.
 Fig. 3 Difference between VDI and Partially Virtualized Local Disk

Table 4. Comparison between VDI and PVLD methods

Topic VDI PVLD

Processing for enterprise systems Enterprise system server Enterprise system server
Local application processing (e.g. Microsoft Office) VDI server OS Local computer OS
User data storage Data center Data center
Offline device usage No Local applications available
Changes when switching to a new device No (identical from any device) No (identical from any device)

However, maintaining stable operation of all applications
through VDI required tremendous deployment and
running costs. In this chapter, we introduce two methods
to solve this issue: virtualization of only the local disks
of corporate machines, and virtualization of only internet
viewing systems.
3.1 Virtualization of the Local Disks of Corporate
Computers
3.1.1 Background
In recent years, security concerns of storing data in
endpoints, such as data loss and leakage, are becoming
widely held. One existing solution is the use of virtualized
computers on a server, accessible via VDI, with the
actual employee machine’s physical, local drive (hard
drive) inaccessible and thus storing no data whatsoever.
In other cases, folder redirection features are used to
redirect certain regions of a local disk, such as the “My
Documents” folder, to a centrally managed server.
However, because VDI centrally hosts both data and all
program execution on the server, the servers are required
to have extremely high processing power, resulting in
extremely high costs for deployment, operation, and
maintenance.
Meanwhile in folder redirection, a user can access both
regions that will be redirected to the server and those
that will not. This means that any data stored in the latter
will not be centrally managed in the server.
In order to remove users’ data from local disks and
centralize that to a server without the high cost
requirements of total VDI, Kyushu Electric developed a
method to restrict data access and virtualize only the local
disks of corporate computers and successfully deployed
this to all 15,000 corporate machines. We call this the
“Partially Virtualized Local Disk” (PVLD) method.
3.1.2 Configuration of the partially virtualized local
disk method
We will now discuss and compare the configuration and
characteristics of a PVLD deployment against a classic
VDI environment.
In a VDI environment, a VDI server is used to process
users’ programs while user data is stored in a data center.
Endpoints are solely used as terminals that display the

Cigre Science & Engineering • N°13 February 2019

148
 Fig. 4 Processing Access in a PVLD

Fig. 5 Step-by-step process of write control

output of the server’s processes. In a PVLD environment,
user programs are all processed on their own local
machines, and only the storage regions for their personal
data (called “Exclusive Personal Storage Regions”) are
hosted on a data center. (Fig. 3)
These structural differences allow user data to be stored
in a data center and accessed from any device like a VDI
environment while using the operating systems on the
computers.
In addition, unlike VDI environments where a computer
is merely a display and thus nonfunctional without
connectivity, external storage or other exceptions can be
used in a PVLD environment to temporarily allow the use
of local applications such as Microsoft Office (Table 4).
3.1.3 Mechanisms of the partially virtualized local disk
We will now illustrate the mechanisms involved in
a PVLD environment. Access to the computer and
Exclusive Personal Storage Region is processed in the
steps listed below and in Fig. 4.
• Accessible and non-accessible regions on the local
disk are defined for each process that a user can operate.
• The following write controls are applied to a user-
operated process attempting to access the local disk.
(a) For data access to accessible regions by a user’s
process, folder redirection points the access to that user’s
personal storage region.
(b) For data access to non-accessible regions, the access
is denied.
Fig. 5 shows a flowchart of actions in a write control
process in a PVLD environment. Detailed explanations
for each step can be found below.
• After the user logs in (step 1), the computer tests to
determine if it can access the server (step 2). If yes,
moves on to step 3; if not, moves on to step 10.
• Step 3: Receives and applies the newest policies for
write control.
• Step 4: Prepares folder redirection toward user A’s
exclusive personal storage region when an
accessible region is accessed.

Cigre Science & Engineering • N°13 February 2019

149
• Step 5: The computer determines if a process was
run by the user. If yes, it proceeds to step 6; if not, the
process is allowed to function normally.
• Step 6: Determines if the user’s process is attempting
to write to an accessible or non-accessible region. For an
accessible region, access is granted and redirected (step
7); for a non-accessible region, access is denied (step 8).
• Step 9: Determines if the user initiated a logout. If yes,
proceeds with the logout; if not, return to step 5.
• If the computer cannot access the server in step 2, access
control grants access only to a limited region (e.g. USB
memory) (step 10) until logout (step 11). A user would
be able to perform some tasks, such as taking notes or
giving a presentation at an external, offline environment.
There is the Active Directory [7] as a tool to restrict
access to inaccessible regions on a local drive. But the
specificity of the restrictions is limited to individual users
or groups and for entire devices or regions. This cannot
give each user granular controls for individual programs
or specific workflows, making PVLD incompatible with
some programs and workflows. This raises the need for
a more versatile alternative solution. Since denying all
access to non-accessible regions would break programs,
whitelisting and blacklisting were combined to make an
access control policy with consideration to processes
and regions (locations) that the user would use.
3.1.4 Benefits of the Partially Virtualized Local Disk
method
Kyushu Electric achieved the following benefits from
using a PVLD environment for all employee computers.
(1) Security Improvements
Users’ data are not stored on the local disk. Lost
or stolen computers are not at risk of data loss and
leakage, resulting in a high level of security.
Limiting Exclusive Personal Storage Region access
to the user’s processes also prevents ransomware and
other malware or cyberattacks that delete or rewrite
personal storage region data. (2) Cost Reduction
Had Kyushu Electric proceeded with their VDI-based
thin client deployment, running all 15,000 corporate
computers as virtual machines on a data center would
have required hundreds of physical servers and one
petabyte of high-performance storage, plus licenses for
Cigre Science & Engineering • N°13 February 2019
150
the VDI solution. The PVLD solution required little
more than storage hardware for the Exclusive Personal
Storage Regions, greatly reducing the involved costs.
(3) Effective use of existing resources
Kyushu Electric had already implemented VDI for
some machines but could not achieve the same level
of performance as standalone machines. Demanding
tasks such as Excel macros caused especially
significant dips in performance, generating complaints
from users. Because PVLD leaves processing to the
endpoints’ hardware resources, performance remains
consistent and complaints have subsided. This makes
PVLD a logical solution that keeps data off endpoints
while making full use of their resources.
(4) Convenience
A user can log into any computer that has been set up
for PVLD and see the same desktop and use the same
applications and data. If a normal computer fails, any
data saved on the local disk is lost – however, with
PVLD, all data is stored at the data center. These
advantages led to increased convenience for users.
Immense amounts of data are being stored and collected at
all times and places in this era. For any organizations, data
loss prevention is not only an internal issue but a matter
of maintaining trust from the people. Unlike VDI, PVLD
effectively utilizes the hardware resources in computational
devices to achieve high security at a low cost.
This method can be perceived as a middle way between
typical standalone clients and total VDI, and we believe
that it will become a new standard for all organizations
of all industriesa, electric or otherwise.
3.2 Virtualization of internet browsing system
3.2.1 Background of the virtualizing internet browsing
application
Although there was no infection in J-POWER’s system,
in the past, J-POWER detected malware attempting APT
(Advanced Persistent Threat: a persistent threat using
advanced technology), and to establish a connection
in an attempt to reach a C&C server (Command and
Control server: a server sends command and control to
the infected PC). Both were blocked before reaching in
and out by a firewall.
 Fig. 6 Diagram of partial thin client system virtualizing a web-browsing application
For security enhancement, J-POWER separated out
the internet accessible environment from the internal
network and disabled the internal network’s access to
the internet directly. J-POWER planned to virtualize
its business application and internet browsing system
and centralize the security management for this internet
accessible area.
As it is a high cost to virtualize every business application
likes thin client, J-POWER decided to virtualize an
application in phases and to start from the highest
necessity business application. J- POWER constructed a
partial thin client system starting with a virtualization of
a web browsing application.
3.2.2 Virtualization of the browser application
Requirements and a policy for the construction of a
partial thin client system with the virtualizing internet
browser application are :
Requirements:
a. Virtualization of internet browsing application
should not affect the amount of internet access
before virtualizing.
b. When the user is lead to a malicious web site, it
should not affect PC for business usage.
c. Initial cost and running cost should be reduced.
Policy: Complete the measures within the fiscal year
to respond with cyber-attacks, getting sophisticated, as
soon as possible.
For system integration, J-POWER developed the system
considering the following.
B As existing proxy servers were allocated for
multiple departments, J-POWER centralized the
access from this system to a single proxy server for
standardization of the security operation.
Cigre Science & Engineering • N°13 February 2019
151
C Place the provisioning server for a virtual web
browser to speed up the delivery to the users.
D J-POWER measured the number of users accessing
internet during peak time. From the measurement,
it was 10% of whole users’ so J-POWER sized the
provisioning server to this number for reducing
initial cost. Fig. 6 shows the diagram of partial thin
client system virtualizing web-browsing application.
E For each client devices, J-POWER changed the
method of internet browsing to application streaming.
a. Native applications for internet browsing will run
on a virtual application server.
b. Use a native internet browser on a client device
for accessing web applications on an intranet
system.
F By applying virtualization technology to virtual
application server, there were two benefits.
a. Virtualization of internet browsing system:
J-POWER was able to optimize the operation
management of software version up and patching.
b. J-POWER is able to handle every client’s
application definitely and speedily because J-
POWER could unitarily manage the application
from the center server.
G J-POWER configured the virtual application server
to run on multiple virtual servers on a physical server.
In case of partial system failure, by failover to other
virtual server and configuring a new virtual server
speedily, J-POWER could reduce the service stop
time and did not need to handle all client devices.
3.2.3 Future works
The main issue of this system is the temporary resource
shortage due to concentration of loading. This is caused
by centralization of functions for each client device
to the center server. To resolve this issue, J-POWER
re-examines the load balance and the center server
is restarted everyday to release resources. From a
performance point of view, it takes time to start up the
internet browsing system from PC and the streaming
speed slows down due to heavy loading in the server.
Though it is not affecting the business, J-POWER is
tuning up the system to improve response time.
4. Conclusion
In this report, SDN was applied to an electric power
company corporate network and demonstrated the ease of
operation and reduction of initial configuration time, and a
partially virtualized local disk and internet viewing system
virtualization was shown to be an effective cost reduction
solution for VDI- based thin client environments.
Cigre Science & Engineering • N°13 February 2019
152
5. Bibliography
[1] OpenDaylight, Linux Foundation Collaborative Projects: https://
www.opendaylight.org
[2] NEC, NEC SDN Solutions, Case Studies: http://www.nec.com/en/
global/solutions/sdn/case/index.html
[3] VDI: Main page: https://en.wikipedia.org/wiki/Desktop_
virtualization#Virtual_desktop_infrastructure
[4] Open Networking Foundation: “OpenFlow Switch Specification
Version 1.3.0 (Wire Protocol 0x04) ”, 2012
[5] OpenDaylight, VTN: Main page: https://wiki.opendaylight.org/
view/VTN:Main
[6] OpenDaylight, Controller: Main page: https://wiki.opendaylight.
org/view/OpenDaylight_Controller:Main
[7] Microsoft, “Active Directory Architecture”, TechNet, 2017. https://
technet.microsoft.com/en-us/library/bb727030.aspx