Anda di halaman 1dari 127

IEEE Power & Energy Society TECHNICAL REPORT

May 2007 PES-TR13


Formerly TP180

Interconnected Power
System Response to
Generation Governing:
Present Practice and
Outstanding Concerns
PREPARED BY THE
Power System Dynamic Performance Committee
Power System Stability Subcommittee
Task Force on Large Interconnected Power Systems
Response to Generation Governing

IEEE 2013 The Institute of Electrical and Electronic Engineers, Inc.


No part of this publication may be reproduced in any form, in an electronic retrieval system or otherwise, without the prior written permission of the publisher.
THIS PAGE LEFT BLANK INTENTIONALLY
IEEE TASK FORCE ON
LARGE INTERCONNECTED POWER SYSTEM RESPONSE TO
GENERATION GOVERNING

Chairmen: Richard Schulz and Pouyan Pourbeik


Secretary: James Feltes

Members and Contributors

Baj Agrawal AI DiCaprio Stefano Massucco


Eric Allen Carlos Grande Edward Miska
Tim Bartel Les Hajagos Nalin Pahalawaththa
Navin Bhatt Howard lilian Mahendra Patel
Terry Bilke Jim Ingleson Shawn Patterson
Elmer Bourque John Kehler Les Pereira
Roy Boyer Rob O'Keefe Alex Schneider
Terry Crawley Leonardo Lima Marino Sforna
Donald Davies Sture Lindahl John Undrill

11
ACKNOWLEDGEMENTS

The TF is part of the IEEE Power Engineering Society, reporting through the Power
System Stability Subcommittee of the Power System Dynamic Performance
Committee. The Scope was approved in April 2001 by the Power System Stability
Subcommittee, and approved in July 2001 by the Power System Dynamic
Performance Committee. We are truly grateful for the support of our sponsoring
subcommittee and committee.

During the course of this work, conversations among TF participants who also
worked within similar activities by other organizations have contributed to this
work. These other organizations include:

NERC and its NERC Resources subcommittee and Task Force.

CIGRE Task Force C4.02.25, which published the report: "Modeling of


Gas Turbines and steam turbines in Combined-Cycle Power Plants".

CIGRE Task Force 38.02.14, which published the report: "Analysis and
Modeling Needs of Power Systems Under Major Frequency
Disturbances"

These conversations have been fruitful. We thank the members and leaders of
these other groups for their gracious and willing support.

We are truly grateful to James Feltes and Patricia Goodwin for their kind and
diligent help in the formatting of the entire document. Pouyan Pourbeik also
provided help in formatting and pulling together the whole document.

iii
CONTENTS

Chapter 1 - Introduction

1.1 The Scope of Work 1-1


1.2 The Need for Frequency Regulation 1-1
1.3 Blocking of Effective Governing 1-2
1.4 Description of the Issues 1-2
1.5 Behavior of Turbine Generators 1-4
1.6 Effect of Load on Frequency Regulation 1-21
1.7 The Influence of Economic, Market, and Reliability Criteria
Considerations on Primary Governing Frequency Response: ... 1-21
1.8 Example of an Energy Market That Has Established Primary
Frequency as an Ancillary Service - New Zealand's Market for
Ancillary Services 1-27
1.9 Including Demand-Side Participation as a Resource for Primary
Governing Frequency Response 1-29
1.10 Summary 1-30

Chapter 2 - Technical Studies and Results

2.1 Introduction 2-1


2.2 Primary Governing and Frequency Control in ERCOT 2-1
2.3 Nordel Interconnection 2-16
2.4 North Dakota / Minnesota Generator Response and Exposure
- The June 25, 1998 Northern MAPP Disturbance 2-27
2.5 Regression Analysis of The Eastern Interconnection 2-35
2.6 Governor Goals and Characteristics in UCTE 2-38
2.7 Development of the New Thermal Governor Model in the WECC
............................................................................................................ 2-47
2.8 New York Observations of Generator Governing Response 2-71
2.9 Summary 2-78

Chapter 3 - Conclusions and Recommendations

3.1 Conclusions on Present Practice 3-1


3.2 Recommendations 3-4

References

Appendix A - List of Acronyms and Terminology

iv
INDEX OF AUTHORS

Authors Listed in Alphabetical Order

Chapter 1 Introduction
B. Agrawal, E. Bourque, L. Hajagos, H. lilian, N. Pahalawaththa, M. Patel,
S. Patterson, P. Pourbeik and R. P. Schulz

Chapter 2 Technical Studies and Results


E. Allen, T. Bartel, T. Bilke, R. Boyer, J. Ingleson, S. Lindahl, L. Pereira,
P. Pourbeik and M. Storna

Chapter 3 Conclusions and Recommendations


E. Allen, R. Boyer, J. Feltes, C. Grande, H. lilian, J. Ingleson, R. O'Keefe,
P. Pourbeik, A. Schneider, R. P. Schulz and M. Storna

Appendix A - List of Acronyms and Terminology


Main Editors: P. Pourbeik and R. P. Schulz

v
1 INTRODUCTION
1.1 The Scope of Work
This report, prepared by the IEEE Task Force on Large Interconnected Power
System Response to Generation Governing (the TF), completes Task 1 and 2 of
the TF Scope. The TF scope was as follows:
Scope: The Task Force (TF) will complete two tasks:
1. Review and document the present practices for generating unit
speed/load controls and the resulting governing (or primary frequency
control) of power systems, including recommendations for modeling
within large interconnections.
2. Consider and document recommendations for appropriate follow-on
activities in the evolving electric utility re-structuring environment.
The TF coordinated its activities with other industry bodies, including those of IEEE,
CIGRE [1], and NERC.

1.2 The Need for Frequency Regulation


The main reason for good frequency control in any system is to support stable
and reliable system operation. System break-ups are rare, however, when they
do occur islands of load and generation are formed. The generation-load levels
in most islands will not be balanced, since power exchange among these newly
formed islands (which occurred while the system was still fully intact) is lost due to
system break-up. Thus, without adequate and proper means of frequency
regulation these islands are prone to collapse. In addition, for such extreme
conditions under-frequency load shedding is also needed. Although
sophisticated schemes for arresting and reversing rapid frequency changes by
under-frequency load shedding (UFLS) have been developed, UFLS is not a
substitute for governing response. Both UFLS and primary governing are needed
and must be coordinated. If generating plants are not well behaved under
conditions of significant off-nominal frequency operation and if they are not able
to provide proper frequency regulation, the consequence can lead to a
blackout of that island (see section 2.6.3). There are many other reasons for
maintaining tight control of system frequency. Among them are the following [2]:
The efficiency and performance of most large power plants is dependent
on the performance of the auxiliary systems. Many of these auxiliary
systems (feedwater pumps, fans, etc.) are motor load driven and thus
their performance is tied to the speed of the motor, which is dependent
on system frequency'.

1 For example, in the case of PWR nuclear reactors, prolonged system under-frequency
conditions can lead to proportional reduction in reactor coolant flow due to a reduction
in the speed of the reactor coolant pumps. This can thus lead to what is known as a
departure from nucleate boiling (DNB). To avoid a DNB condition, reactor under-
frequency trip settings are provided based on the reactor coolant power supply

1-1
Excessive drop in system frequency can result in considerable increase in
the magnetizing current drawn by induction motors and transformers.
The integral of system frequency is used by many electronic clocks and
other devices for the purpose of keeping time.
Excessive off-nominal frequency operation (up or down) can lead to
increased fatigue on turbine-generator blades (see Table 2-4, in section
2.4). In fact, most turbine-generators will trip if frequency deviates from its
nominal value by more than + j- 5%.
For these reasons, system frequency must be tightly controlled at all times.

1.3 Blocking of Effective Governing


Owners/operators of generating units have strong economic reasons to operate
generating units in many ways that prevent effective governing response. If a
unit is operated at its full megawatt capability, it will not be able to increase
power in response to a decrease in system frequency, regardless of the unit's
control mode(s). A unit operated with any of several control modes or settings
will be unable to effectively provide governing to either upward or downward
changes in frequency. Some examples are outer-loop megawatt control (that
slowly adjust the governor load reference to maintain unit output at a pre-
selected megawatt level) and intentional deadbands or limiters in the unit's
control. These issues are discussed in more detail throughout this report.

1.4 Description of the Issues


In recent years, power system operating and planning personnel have become
increasingly aware of the fact that power plant governing response is
considerably less than expected and planned [4].
This issue was addressed, from the operations side, by the (then) NERC Operating
Committee and its Performance Subcommittee, who initiated an EPRI project [5].
Personnel in the system operating components of utilities conduct 'regulation
tests' that are initiated by and coordinated by the NERC Operations
Subcommittee. The committee reached a broad consensus that approximately
one-quarter to one-third of the expected governing response is found in analyses
of the recorded power system frequency. Similarly, recent work in the Western
Electricity Coordinating Council (WECC) [6] also investigated the discrepancy
between previously used models (which overestimated the primary frequency
response of the system) and real measured events. Through their work it has
been shown that the discrepancy is primarily due to non-responsive governor-
turbines (see Chapter 2, section 2.7).
From the planning side, the problem is that power system analysts, using
conventional transient stability program methods for simulating interconnected

frequency. In a 60 Hz system such as the US, a typical under-frequency trip setting for this
condition is 57 Hz [3].

1-2
electric utility system response, are consistently unable to replicate the recorded
response of the power system to loss of a unit and its megawatt output. The
difference between simulated and recorded system frequency response to the
loss of a unit has these characteristics:
a) The simulated rate of fall of system frequency following unit trip
matches recordings. This indicates that the modeled value of total
inertia, H, is valid.
b) In some studies, the oscillatory component of the simulated
response of system frequency matches the recorded oscillatory
response of system frequency [7]. This indicates that the modeled
generation-load conditions and the modeled transmission system
match the power system conditions. In other studies, the simulated
response of system frequency does not match the recorded
oscillatory response of system frequency [6]. This oscillatory
response difference has been attributed to differences in the
modeled dispatch of generation and initial loads and interchange.
c) The sustained component of the simulated response of system
frequency is closer to 60 Hertz than the recorded sustained
component of system frequency. This difference is "the governing
problem".
There is an apparent consensus (among personnel within electric utility
transmission system operations, and some transmission system planners and
power generating personnel) that the present governing response is
considerably less than the governing response that is set in power plant
governors. This is the disparity under consideration. The response is
approximately one quarter to one third of the response that would be expected
from governor settings. There is a further consensus among the same personnel
that accepts the present governing response as: a) adequate for security and b)
an appropriate economic tradeoff under normal operating conditions - it must
be emphasized that this latter statement is in the context of the North American
system, the world's largest power system. In smaller systems, such a disparity
between actual and expected primary governing could be quite disastrous. It
should be added that current trends indicate a declining primary governing
response, and that if the declining trend continues, all interconnections will
eventually be put at risk for inadequate security.
Conventional modern power systems use three phase alternating current (ac)
generation and transmission equipment that innately operates in synchronism. A
key feature of synchronously operated power systems is that power flows
change immediately to make the entire generation of an interconnection to
support the entire interconnection load. Consequently, synchronous operation
permits sharing of generation reserves. This design feature is both economical
and reliable since: a) ac operation offers the economy of high voltage
transmission with its much lower losses; and b) shared generation reserves allow
each utility to build and operate with less generation while improving reliability.

1-3
Each of these two economic features is worth millions of dollars per year for each
utility.
Synchronous operation among utilities also offers the ability to buy and sell
energy over the transmission system, with added economic benefits. However,
utilities must regulate their generation to control the commitments to provide
power. This generation control must have provisions to allow each participant in
the energy market to match the generated power to meet the sales of electric
energy.
System frequency and interchange power levels are the natural indicators of the
balance between power production and energy commitments, since
synchronous operation will naturally and immediately shift power transfers to
distribute power generation to power consumption. As interconnection-wide
imbalances between power production and power consumption continue, the
interconnection frequency will rise or decline.
Utilities use these indicators (system frequency and interchange power levels) to
buy or generate energy to: a) maintain a balance between power production
and energy commitments; and b) to replace generation that is lost for
unintended reasons. These are fundamental roles for power system control
centers, including governing and automatic generation control (AGe).
In summary, the key points here are that governing is a primary function within
interconnected (and islanded) power systems. In many power systems today,
there is a discrepancy between actual system frequency response to a
generation/load imbalance event and what is predicted using standard power
system simulation tools and simulating the event based on the expected
governing response of units. This report is devoted to providing insight on what is
the cause of this discrepancy and how more refined modeling practices may be
adopted to bridge this gap between actual and simulated system response.

1.5 Behavior of Turbine Generators


1.5.1 Fossil-Fuel steam Power Plants
Fossil-fuel steam power plants generate electrical power by boiling water to high
pressure and temperature, and then allowing the steam to expand through a
steam turbine that drives the electric generator. Most large steam turbines will
have multiple sections (high pressure, intermediate pressure and low pressure)
that may be connected either in a tandem compound (on a single shaft) or
cross-compound configuration (multiple shafts). Each shaft will have its own
electric generator. There are two categories of steam turbines. Drum-type
boilers that boil water to generate a steam vapor mix and separate the vapor
from the liquid in the boiler drum. The second type is a once-through design,
where water coming into the boiler is raised to supercritical pressures, usually
above 3500 psi, where there is no identifiable gas or liquid phase of the water.
Steam power plants constitute the largest portion of the generation mix in many
regions. They are used for base load, for daily and short term load following and
for frequency regulation. When base loaded they are controlled to not

1-4
contribute to primary frequency control for an event resulting in a decline in
system frequency, although the steam turbine governor will still respond to over-
frequency events. Like gas turbines and modern combined-cycle power plants
(see section 1.5.4), if operated on outer-loop megawatt control, they may
contribute initially to primary frequency regulation, following a frequency
excursion, however, the megawatt control loop will then bring the unit back to its
original megawatt output set point and thus defeat the units droop response
(see Chapter 2).
There are four possible control modes that the steam turbine may be operated
under while on speed-governor control and able to provide primary frequency
control based on its governor droop settings. The four possible control modes
are (i) boiler-follow, (ii) turbine-follow, (iii) coordinated control or (iv) sliding
pressure control [2].
In boiler-follow mode the turbine main control valves control the power
generated, and the boiler is controlled to maintain steam generation to meet
the turbine demand. Boiler-follow mode is common for drum units, but rarely
used in once-through supercritical units. Following a demand for increase in
power a burst of power can be provided by quickly changing the steam flow to
the turbine. However, as the control valves are opened, the subsequent
increase in steam flow results in a steam pressure drop and thus a drop in power
generated. This pressure change is then typically used as a feedback signal to
increase fuel (pulverized coal) and air flow into the furnace, increasing steam
generation and pressure. Thus, the turbine response under this control scheme is
characterized by a prompt and sustained increase in power followed by a long
delay (tens of seconds to minutes) until the full increase in power is achieved as
steam pressure is restored. One of the potential problems with this approach is
that the initial, and sudden, response of the turbine control valves to a request
for additional power, if excessive, may result in enough of a pressure transient as
to destabilize the boiler "drum level and thus result in a unit trip. Ensuring proper
control design, to maintain drum level, typically avoids this problem.
In turbine-follow mode (typically used for once-through units) the response of the
unit to increases in power demand is achieved through controlling the fuel
source into the boiler. The turbine main control valves are used to maintain
constant steam inlet pressure. Thus, the response of a turbine-follow unit is
naturally slower than that of a boiler-follow unit.
Coordinated control offers a balance between the above two schemes [8]. In
this control scheme the intent is to achieve a compromise between the initially
fast response of a boiler-follow design and the more sluggish response of a
turbine-follow control, while maintaining much better control of the boiler than
pure boiler-follow control.
The final approach is sliding-pressure control. In this control mode, the main
turbine control valve is left wide-open and power is changed by changing the
steam throttle pressure reference as a function of turbine power output, rather
than keeping it at a constant setpoint. This is true provided the drum steam
pressure is above a minimum setpoint. Once throttle pressure falls (at relatively

1-5
low loads) the control valves begin to close. Sliding-pressure control achieves
greater overall turbine efficiency by maintaining steam inlet temperature into the
high pressure (HP) turbine essentially constant. Sliding-pressure control is typically
done on steam turbines in combined-cycle power plants (see section 1.5.4). As
in turbine-follow control, this control mode will result in a much more sluggish
response of the unit to primary frequency regulation since the demand for
increased power (for a frequency decline event) is effected through controlling
the boiler and fuel, which respond much more slowly than do the turbine control
valves.
1.5.2 Hydro-Turbines
1.5.2. 1 Hydro Response
Hydro governors are essentially no different than their thermal unit counterparts.
Neglecting, usually minor, effects as deadband (either intentional or
unintentional) any speed governor controlling a generator will respond in a very
predictable manner to measured speed changes and alter flow to the turbine.
The response of a hydro generator unit to system frequency changes can be
greatly impacted by turbine characteristics and required compensation for the
inertia of the water column and control loops external to the turbine-governor-
generator system. The operation of a hydro plant tends to be simpler than
thermal facilities and less constrained by other plant processes or concerns.
Therefore, the response of a hydro unit to system frequency deviation is fairly
reliable, although the exact amount of response can vary depending upon the
turbine and the unit operating conditions.
As the results of system wide measurements made during staged testing in the
WECC system [6] showed, hydro units are normally quite responsive to a sudden
frequency drop, provided the units are loaded sufficiently below their maximum
output (This observation that most hydro units respond to a drop in system
frequency is also corroborated by the observations made by the NYISO, see
section 2.8). In general, due to the compensation of the inertia of the water
column the power response rate of a hydro unit is relatively very slow compared
to thermal units. The governor is normally adjusted to provide a stable response
when operating under isolated load conditions. This results in a slowed power
response rate when operating in a large, synchronous system [9], [10], [11]. The
resulting time constant of the generator-governor-turbine system is, typically
greater than 30 seconds. Operating with the dashpot bypassed allows a much
quicker power response of units with mechanical governors, but they will be
unstable in islanded conditions. Modern digital control algorithms can switch
controller parameters when islanding is detected [12], allowing for optimal
response under any system condition. However, these units are still rare.
Simulations of frequency disturbances in the WECC system show that, on
average, the hydro units will achieve two-thirds of their final value in about 60
seconds. Therefore, on average, hydro units will impact the transient dip in
frequency after a generation loss only minimally, completing about 25 percent
of its total response in the first 10 seconds. Complete response is typically in the
timeframe of one to three minutes.

1-6
Over the wide range of possible operating conditions, the power output of a
hydro turbine changes nonlinearly with turbine flow. Turbine flow is further
dependent upon other factors, such as plant head, turbine speed, and penstock
characteristics. Operating points and capacities of units in plants with reservoirs
can vary widely from season to season due to head variation. While hydro units
are usually operated in a mode which tends, on average, to somewhat
decrease the nonlinear effects, in their more extreme ranges these effects can
result in responses significantly different than expected for a given governor
action.
In order for generators to achieve a more predictable and equitable proportion
of their power response, an electrical power signal might ideally be used for the
feedback signal. However, for several valid reasons, most older hydro governors
utilize wicket gate position as a surrogate for the power signal. The resulting
nonlinear gain relationship between gate position and electrical power output
accounts for most of the variability in the response of a hydro unit. Some newer
digitally controlled systems are designed to provide a constant, power based
droop effect, but these systems will likely be in the minority for some time.
The relationships between gate position and power delivered by the turbine
varies with the generator operating point and can be fairly unique from turbine
to turbine, but share some general characteristics. Figure 1-1 illustrates some
example relationships between gate position and electrical power for a few
different hydro units. Also shown is the 1:1 slope that assumes that gate position is
always equal, in percent of rated, to electrical power output. The plot illustrates
that at most operating points, the relationship will have a slope other than 1.0,
and therefore, for most operating points the response to system frequency
deviation will not have a final power response that equates to the governor
movement (droop setting). Typically, over the lower and mid-range of operating
points, the additional gate/power turbine gain is around 1.5 per unit, which will
result in a more responsive governor, and a speed regulation (sometimes
referred to as effective droop) value of 3.3 percent for a nominal droop setting
of 5 percent. At operating levels above 90 percent, the gain tends to taper off
below unity, and the unit will be less responsive than expected. Some available
simulation models take these effects into account.

1-7
..!

30
40.._SO..6t
Gate Posmon percent
w
10w..
,...,J
80 90
100

Figure 1-1. Examples of Gate Position vs. Power Characteristics for Hydro Turbines

This nonlinear relationship is due primarily to the fact that the hydro turbine
efficiency changes, often quite dramatically, with the flow of water through the
turbine. In general, hydro turbines tend to be very inefficient at lower flow levels
and have the highest efficiencies at flow levels approaching the rating of the
turbine. They are usually designed for optimal efficiency somewhere in the range
of about 75-90 percent flow (gate position.) Therefore, as the gate position
increases towards or decreases from this range, so does the efficiency.
Consequently, the gain from gate position to power output varies with the
operating point. This gain is typically the greatest between 40 and 80 percent
gate position. If a unit is operating at peak efficiency or above, an increase in
gate position will include a decrease in turbine gain (the Power vs. Gate curve
will result in a smaller increase in output power) while if a unit is operating below
peak efficiency, an increase in gate position will be met with an additional gain
from the turbine (the Power vs. Gate curve will result in a larger increase in output
power.)
Of course, like a thermal unit, a hydro unit that is loaded to maximum power
output will not be responsive to decreases in system frequency. As a rule,
multiple unit hydro facilities are not usually loaded in this manner, but rather
dispatched to provide peaking power and/or spinning reserve . However, in
extreme water supply conditions or cases of insufficient unit availability, units may
be run at full output to maintain a critical water flow. Many, smaller, often single
unit plants may be relied on primarily for water flow control, and may not be
responsive to system frequency at all. In most cases, these "run of the river"
plants should not be considered frequency responsive.

1-8
Also like thermal plants, hydro units may be under the influence of outer control
loops. A plant may be responsive to a control area setpoint which maintains a
total plant output and resets any power changes initiated by the individual
governors. In some plants, there may be plant coordinated control systems,
designed to maintain a flow schedule or maximize efficiency. If not properly
biased to allow governor-generation response to frequency changes, these
controllers may interfere with the regulation of system frequency.
1.5.2.2 Hydro Turbine Modeling
In the past, the modeling of governors and prime movers in large scale system
simulations was given little attention. Studies were limited to the transient (first
swing) or very short-term time frame. These models were simple approximations
designed to adequately represent the governor and turbine in this short time
frame, without unnecessary detail. In time, detailed models of the governors
were added [13] to allow tuning of governors for small-system stability [14]. These
available models are entirely adequate for representing the dynamics of hydro
governors and their accuracy in modeling the control of wicket gate opening
and closing is easily validated.
The simple, linear approximation of the hydro turbine and water column,
however, is still in wide use. The model shown in Figure 1-2a has been used in
simulations since the simple governor model was introduced. It is a useful
approximation for many purposes, but falls short of adequacy for studies
involving large or sustained frequency deviations.
One of the problems inherent in using the classical, linear model is that it does
not automatically account for the effects of changing flow, speed, etc.
Although some versions of this model allow setting all of the partial derivative
based parameters, they need to be manually reset for simulation at different
operating points and will not change dynamically. These models are almost
always set with parameters that assume linear operation at full load, so the
model simplifies to Figure 1-2b. Using this model at different loading points will
make the dynamic response inaccurate, since the effective water starting time
constant, TW, varies in direct proportion to gate position. The dynamic
characteristics of a hydro unit can be very dependent on the loading level.

Power . .
...

a. Generalized, Operating Point Dependent Model

Gate Power ...


~ ...
Position

b. Classic Model with Ideal Assumptions

Figure 1-2. Turbine/Water Column Models

1-9
A significant amount of error will arise if the linear turbine models are used for
post transient. mid-term, or longer simulations involving changes to the
generation/load balance. The nonlinear turbine effects will strongly influence the
overall turbine gain and the amount of power response.
Many very detailed hydro turbine models have been developed and used, to
various but limited extents [15], and generally only are necessary in very detailed
studies of single plants. An example of a model that achieves a compromise
between simplicity and capturing essential behavior for large- scale studies is
shown in Figure 1-3. This type of model offers some improvement on the classical ,
linear model since it is capable of dynamically altering the effective time
constant. Tw, in addition to including additional gain modifications due to turbine
damping, head variation, and most importantly, the nonlinear relationship
between gate position and mechanical power.

1
T S
Pos ition

Figure 1-3. Nonlinear Turbine Model


Such a turbine model is appropriate for most purposes for Francis turbines,
impulse turbines, and fixed blade propeller turbines. However, if variable p itch
blade Kaplan turbines are to be modeled, there is an additional control loop
that is not included in these models which is essential to obtaining the correct
response in simulation [16].
The flow dependent, nonlinear behavior of the hydro turbines can significantly
impact the hydro unit response to a deviation in system frequency. The resulting
gain (and effective droop) of a hydro governor reasonably can be expected to
vary as much as 50 percent over the operating range of the unit. Therefore, it is
evident that extra care must be taken in developing how generation is
dispatched in the load flow model, ensuring that hydro models are realistically
loaded.
Hydro units that operate with fixed gate pos itions or that regulate flow rather
than speed should be represented in simulation w ithout any governor model.
Similarly, in most cases, governor models should not be attached to hydro units
that are used as synchronous condensers. These units may be considered
operationally as spinning reserve, but in general are not responsive to system
frequency, since their reference adjusters and/or gate limits are typically set so
the wicket gates will not open even for the larg est frequency changes.
Exceptions to this rule are digital systems which have been specifically designed
to switch from condensing mode to frequency responsive mode.

1-10
It should be identified when a hydro unit is under the influence of an outer loop
controller so the effects can be added into the model to inhibit or counteract
the governor response as appropriate. It is likely that the actual controllers can
easily be modified to allow a temporary free response of the governor during
system frequency disturbances. This would be a better solution.
1.5.3 Nuclear Units
1.5.3. 1 Pressurized Water Reactor
A pressurized water reactor (PWR) is a type of nuclear power reactor that uses
ordinary light water for both coolant and for neutron moderator.
In a PWR, the primary coolant loop is pressurized so the water does not boil, and
heat exchangers (steam generators) are used to transmit heat to a secondary
coolant which is allowed to boil to produce steam either for warship propulsion
or for electricity generation. In having this secondary loop the PWR differs from
the boiling water reactor (BWR), in which the primary coolant is allowed to boil in
the reactor core and drive a turbine directly.
PWR is the most common type of nuclear power reactor. More than 230 are in
use worldwide to generate electric power, and several hundred more for naval
propulsion. The design originated as a nuclear submarine power plant.
A typical PWR has fuel assemblies of 200-300 rods each, and a large reactor
would have about 150-250 such assemblies with 80-100 tonnes of uranium in all. It
produces electric power in the order of 900 to 1500 M W.
A key mechanism that controls any nuclear reactor is the rate at which fission
events release neutrons. On average, each fission releases just over two neutrons
with a lot of heat. When a neutron strikes a uranium atom a further fission event
can occur, and this can lead to a chain reaction. If all neutrons were released
instantaneously, their number would grow very fast, resulting in the destruction of
the fuel cells and a melt-down of the reactor. However, a small fraction of these
neutrons are released over an extended period (perhaps one minute). This small,
but crucial, delayed release permits the other control mechanisms (negative
temperature coefficient, human or computer manipulation of neutron-absorbing
control rods, etc.) to have an effect.
In PWRs the primary coolant loop flow is established and maintained by the
reactor coolant pumps (RCP), to properly cool the nuclear core. There are
multiple RCPs. All RCPs are needed to support full unit power level. To safely
maintain full core flow, the core flow is not controlled by valves. Each RCP
impeller is driven by a dedicated induction motor. As RCP power supply
frequency drops, the reactor core flow drops accordingly. Some PWR units have
motors, pumps, piping, and reactor core flow passages that are adequate at
rated system frequency, but at lower supply frequency and lower reactor core
flow portions of the core lose heat transfer capability in a process known as
departure from nucleate boiling (DNB). DNBR is a serious limitation on safe
operation, and these PWR plants have under-frequency relays installed to trip
the unit before DNB occurs [3], [17], [18].

1-11
1.5.3.2 Boiling Water Reactor
A boiling water reactor (BWR) is a light water reactor design used in some
nuclear power stations. It has many similarities to the pressurized water reactor,
except that in a BWR the steam going to the turbine is produced in the reactor
core rather than in a steam generator or heat exchanger. In a BWR there is only
a single circuit in which the water is at lower pressure than in a PWR (about 75
times atmospheric pressure) so that it boils in the core at about 285C. The
reactor is designed to operate with 12-15% of the water in the top part of the
core as steam, resulting in less moderation, lower neutron efficiency and lower
power density than in the bottom part of the core.
Reactor power is controlled via two methods: by inserting or withdrawing control
rods and by changing the water flow through the reactor core.
Positioning (withdrawing or inserting) control rods is the normal method for
controlling power when starting up the reactor and operating up to
approximately 70% of rated power. As control rods are withdrawn,
neutron absorption decreases in the control material and increases in the
fuel, so reactor power increases. As control rods are inserted, neutron
absorption increases in the control material and decreases in the fuel, so
reactor power decreases.
Changing (increasing or decreasing) the flow of water through the core is
the normal method for controlling power when operating between
approximately 70% and 100% of rated power. As flow of water through the
core is increased, steam bubbles ("voids") are more quickly removed from
the core, the amount of liquid water in the core increases, neutron
moderation increases, more neutrons are slowed down to be absorbed
by the fuel, and reactor power increases. As flow of water through the
core is decreased, steam voids remain longer in the core, the amount of
liquid water in the core decreases, neutron moderation decreases, fewer
neutrons are slowed down to be absorbed by the fuel, and reactor power
decreases.
1.5.3.3 CANDU Reactor
The CANDU reactor is a pressurized-heavy water, natural-uranium power reactor
designed in the 1960s by a partnership between Atomic Energy of Canada
Limited and the Hydro-Electric Power Commission of Ontario as well as several
private industry participants. CANDU is a registered trademark and stands for
"CANada Deuterium Uranium". All current reactors in Canada are of the CANDU
type. Canada markets the power-reactor product abroad.
1.5.3.4 Governing Characteristics of the Nuclear Plants
Nuclear units are generally operated as base loaded units with load limiting. This
is to avoid exceeding the maximum reactor output to beyond what they are
licensed by Nuclear Regulatory Commission. Unfortunately, load limiting is very
widely used even when the unit is operating at less than its maximum capability.
Thus, a nuclear unit would typically not respond to a system under-frequency
condition except for the initial response due to inertia. This has been seen in

1-12
numerous recordings. Figures 1-4 below shows a typical system under-frequency
event and the corresponding response of a PWR unit (Figure 1-5).

Frequency

6001
6000
5999
[;- 5998
~ 5997
@ 5996
---
I
~

~ 5995
\.-,
.g;

V
5994
OJ 5993
5992
599 1
5990
10:54:03 10:54:20 10:54:37 10:54:55 10:55:12 10:55:29 10:55:47 10:56:04 10:56:21
Time

Figure 1-4. Typical System Under Frequency Event

NET-MvV

1302
1300
1298
1296
1294
1292
1290
10:53:46 10:54:29 10:55:12 10:55:55 10:56:38

Figure 1-5. PWR Nuclear Unit Response to Event shown in Figure 1-4

However, a nuclear unit would typically respond to system over-frequency


conditions and reduce its output. Figure 1-6 and Figure 1-7 below shows the
recorded system frequency and the corresponding response of a PWR unit for a
recent over-frequency event.

1-13
Frequency for 9/12/05 Event

60.25 , . . . - - - - - - - - - - - - - - - - - - - - - - - - - - - - ,

60.2 +-- - - - -- -- - - -- - - - - - - - - - -- - - 1

~ 60.15 +-- - - - - - - - H- - - - - - - - - - - - - - - - --1


>.
g 60.1
Ql
~
C'"
e
u.
60.05

60

59.95 - . - - - - - - - - - - , - - - - - - - - , - - - - -- - - - - - - 1
o 100 200 300 400 500
Time (Sec)

Figure 1-6. Typical System Over-Frequency Event

Unit Net MW for 9/12/05 Event

1395
1390
1385
1380
In 1375
==III:=
III
1370
Cl

::!: 1365
Ql

1360
1355
1350
1345
0 50 100 150 200 250 300 350 400 450 500
Time (sec) Starting at 20:05 UTe

Figure 1-7. PWR Nuclear Unit Response to Over-Frequency Event shown in Figure 1-6

1-14
1.5.3.5 Governing with a Nuclear Unit in Canada
New Brunswick Power operates in the Maritime Area (6000 MW on winter peak)
of NPCC, which is attached to the Eastern Interconnection by a single high
voltage transmission line. This line historically trips about every 1.5-2 years,
resulting in a Maritime 'island'. Point Lepreau Generating Station is a CANDU 600
nuclear power plant located in New Brunswick with a rated output of 680 MW.
When the plant began operating in the early 1980s, the Maritime Area load was
much smaller. One concern was that if the Nuclear Station did not assist in
regulating frequency during power system 'upsets', it might trip due to excessive
system frequency excursions, thereby bringing down the grid. Strong
representation from System Performance and System Operations contributed to
persuading those responsible at the plant and AECB to do as much as possible
to govern on a 4% droop when they could. A droop of 4% is customary in the
Maritime area because of the proven need for good frequency response.
The turbine-governor follows the 4% droop on an under-frequency event until
boiler pressure begins to drop to a minimum accepted limit. The reactor during
the process is at constant power. By the time the turbine power is reduced to
less than the pre-disturbance level due to steam pressure constraints, the other
plants in the power system should be contributing their share of the response. On
over-frequency events, steam is dumped to the condenser or to the
atmosphere, if the event is sufficiently severe, also leaving the reactor at
constant power.
This scheme has proven itself on numerous occasions since 1981, as the Maritime
Area 'islanding' test (shown below in Figure 1-8) demonstrates when the tie to
the Eastern Interconnection was opened for planned maintenance work while
still exporting a nominal amount of power.

1-15
Lepr #1 2006SEP05 1008
580 , - - - - - - - - - - - - - - - - - - - - - - - - - -.........., 2.0

570

560 1.5

550

540 1.0

3: 530 N
J:
:E
I
520
J 0.5

510

500
........-. If 10% Droop
- lf 4% Droop
490
- Lepr #1 MW
dFreq
480
0 10 20 30 40 50
Minu tes

Figure 1-8. Test of Maritime System Islanding and


Subsequent Response of Point Lepreau Generating Station

1.5.4 Combined-Cycle Power Plants and Simple-Cycle Gas Turbines


The recent document [1] gives a comprehensive account of the dynamic
behavior of combined-cycle power plants and how they should be modeled for
power system studies. A brief description of some of the key issues pertaining to
governing performance o f combined-cycle power plants is presented in this
subsection .
A combined-cycle power plant, in its simplest form, consists of a gas turbine, a
steam turbine, a heat-recovery steam generator, and an electric generator. The
combined-cycle plant functions based on the integration of two cycles: the high
temperature (topping) gas turbine (Brayton) cycle and the low temperature
(bottoming) steam turbine (Rankine) cycle. The two cycles a re coupled by
means of a heat exchanger transferring the exhaust heat-energy of the topping
(gas turb ine) cycle to the bottoming (steam turbine) cycle. Thus the name
combined-cycle, since the heat out of the gas turbine exhaust gas is recovered
in a heat recovery steam generator (HRSG) to produce steam for the steam
cycle. The primary advantage of combined-cycle power plants is improved
overall plant efficiency; the latest technologies achieve total plant efficiency
reaching 60%.
A variety of combinations exist, employing multiple gas turbines, HRSGs, and
generators in many possible configurations. These can be categorized into two
main groups:

1-16
Single-shaft units: both the steam and gas turbine share a common
mechanical shaft with a single electrical generator, and
Multi-shaft units: the gas and steam turbines are mechanically
decoupled and each has its own dedicated electrical generator
connected to the turbine on a single shaft.
In both of the above configurations, the total installed generation capacity is
typically in the ratio of two thirds being from the gas turbine(s) and one third from
the steam turbine.
Gas turbines usually consist of an axial compressor, a combustion chamber and
a turbine. These three elements form the thermal block complemented by the
air intake system, the exhaust system, auxiliaries and controls. The steam turbine
is in many ways similar to a typical fossil fuel steam plant, the main difference is in
the way the unit is operated.
Typically, for load levels greater than 50% the steam turbine in a combined-cycle
plant is operated in sliding pressure mode (although fixed pressure control is also
used or often a combination of the two depending on the loading level of the
steam turbine). In sliding pressure control, the throttling or control valves are fully
open and the steam pressure is a function of the steam mass flow entering the
steam turbine. The power output of the steam turbine depends on the steam
mass flow and is not directly controlled. The power output of the steam turbine
can only be increased by increasing the steam flow, which involves generating
more steam in the HRSG by either increasing the heat from the gas turbines or
supplemental firing of the boiler, if present. Therefore, steam units operating in
sliding pressure mode essentially follow the power output of the gas turbines with
a relatively long time constant associated with the boiler (a minute to several
minutes).
The load control and frequency response of a combined-cycle power plant are
handled by the main plant control system. An overall plant load control system
receives a load set-point signal and determines how the gas turbine should be
loaded. As explained above, the steam turbine is generally operated in sliding
pressure mode, thus the electrical output of a combined-cycle power plant is
controlled by the gas turbine only. The steam turbine will follow the gas turbine
by generating power with whatever steam is available from the HRSG.
After a gas turbine load change, the steam turbine load will adjust automatically
with a few minutes delay dependent on the response of the HRSG. There have
been suggestions that independent load/frequency control of the steam turbine
should be provided for sudden load changes. However, such systems would
require the steam turbine to be operated under continuous throttle control,
resulting in significantly lower efficiencies at full and part load conditions.
In order to avoid continuous action of mechanical parts and thus extend the life
of the gas turbines, a frequency dead-band may be introduced in the control
system within which the plant will not respond (for example, in the US, typically a
deadband of 0.025% is introduced into the speed governor control loop) .
Outside this dead-band, a droop setting is followed. The droop characteristic

1-17
setting is set based on grid operator requirements and is in the range of 3 to 8 %,
(typically 4 to 5 % in the US). Due to their relatively high efficiency combined-
cycle power plants are often base-loaded, but they can be operated to supply
frequency support (spinning reserve). For frequency support, the gas turbines
are typically operated between 40 and 95% load, resulting in a proportionate
partial loading of the steam turbine. Simple cycle gas turbines are typically used
as peaking units, that is they are operated for a limited number of hours during
the year when the system is near or at peak load.
A few comments are pertinent about the response of heavy-duty gas turbines,
whether as stand alone (simple-cycle) or part of a combined-cycle power plant.
When base-loaded (Le. at its maximum power output for the prevailing ambient
conditions - the base load of a gas turbine is dependent on ambient conditions
[1]) a gas turbine is under temperature control. That is, the fuel flow in the
combustion process is being limited to maintain a constant turbine inlet firing
temperature (this is often done by control exhaust gas temperature due to the
difficultly of measuring turbine inlet-temperature [1]). When the unit is base-
loaded and on temperature limit, the gas turbine output will not be allowed to
increase by the temperature control loop if a disturbance results in a decrease in
system frequencv-. In fact, for severe off-nominal frequency conditions the
output of the gas turbine will decrease due to the airflow characteristics of the
compressor [1], [19], [20]. In some designs a 'peaking' or 'over-firing' capability is
provided to overcome this phenomenon. In essence, a higher temperature limit
is allowed under severe system conditions to allow the turbine to increase its
output. This, however, has significant financial implications as it reduces the life
span of the hot gas path parts of the turbine and thus increases the
maintenance requirements for the unit. Also, improper design or control under
'peaking' or 'over-firing' operation may lead to the unit tripping due to
overheating right when it is needed most (see Italian experience under section
2.6.3.1 ).
For both combined-cycle and simple-cycle gas turbines, early designs of
combustors and their associated controls were prone to flameout due to sudden
abrupt control commands to decrease fuel flow (e.g. usually such a sudden
demand to decrease fuel flow is due to a response to sudden increase in system
frequency in a small islanded system following, for example, the loss of a
significant portion of load). This is primarily due to sudden transitions through the
numerous combustion modes, resulting in low fuel to air ratio quenching the
flame. Under normal operating conditions such combustion mode shifts take
many tens of seconds. It is incumbent on the host utility/industry and turbine
manufacturer to consult and understand the potential for such transients and to
protect against the possibility of flameout through proper control design (e.g.
effecting rate limits on changes in fuel flow).

2Note: if the initial gas turbine power is below base-load, the turbine power will transiently
overshoot its temperature limit as the temperature controller takes over. Proper design is
needed to prevent the unit from tripping due to excessive temperature transients (see
section 2.6.3. 1).

1-18
Another control feature worth noting is outer loop megawatt control, or
sometimes referred to as 'megawatt pre-select mode'. When partially loaded, a
gas turbine will be under governor control. It will remain under governor control
until it reaches base-load and the temperature control loop takes over.
However, most modern simple-cycle and combined-cycle power plants have an
outer-loop PI-regulator that allows the operator to preprogram a specific
megawatt reference. For example, an operator may be requested to bring a
100 MW gas turbine on-line and to place the unit at 50 MW. To do this, once the
unit has been started the operator may place the unit in 'megawatt pre-select
mode' at 50 MW. This will automatically bring the unit up to 50 MW and maintain
the units output at that level until further instructions are given by the operator.
Under this control strategy, if a system event results in a decline in system
frequency then the gas turbine-governor will initially respond to the event and
increase the turbine output. However, in a matter of tens of seconds
(depending on the controller gain and response time) the outer-loop megawatt
control loop will bring the units megawatt output back to 50 MW - the pre-
selected megawatt output. This has the result of effectively removing governor
action in a short period of time following the event. Reference [6] presents a
comprehensive study performed by WECC to model this behavior not only for
gas turbines but also for all other thermal units. This is also discussed in greater
detail in section 2.7 of this report.
Such 'megawatt pre-select control' will also override the turbine-governors
response to over-frequency events. If left in this mode (or the governor set-point
integrates/winds up while the unit is under temperature limit) the unit could
become non-responsive for significant increases in the system frequency, which
can equally be a problem.
Some small combined-cycle plants are based on aero-derivative gas turbines.
For more detail on these types of units refer to [1]. In general, the main
distinguishing feature between aero-derivative and heavy-duty gas turbines
(from a power system dynamic performance standpoint) is that most aero-
derivative turbines, being multiple-spool- designs, have significantly lighter inertia
as seen by the system (since typically only the low-pressure or "power" turbine
stage is mechanically connected to the electrical generator). Also, aero-
derivative gas turbines are often used in cogeneration applications. In
cogeneration the steam produced by the HRSG is either partially or entirely used
for district heating or industrial purposes. As such, the load level of a
cogeneration plant tends often to be dictated by steam requirements of the
downstream process. Typically, fluctuations in the steam flow and or pressure
can cause significant interruptions in the industrial process using steam and can
thus result in significant financial losses. As such, cogeneration plants often tend
to be considered as intermittent resources since the electrical power output is
often dictated by the steam requirements of the back end process and cannot
be easily dispatched.

3 By multiple-spool design is meant designs where the axial compressor (and turbine) are
split into two or more mechanically separate sections, such that each section runs at a
different rotational speed.

1-19
1.5.5 Wind Generation
The soon to be published document [21] gives a comprehensive account of the
dynamic behavior of wind turbine generators and how they should be modeled
for power system studies. Also, presently the Western Electricity Coordination
Council (WECC) is working on developing generic wind turbine generator
models for use in the most commonly used power system simulation tools in the
USA. The IEEE TF on Dynamic Performance of Wind Power is also supporting the
WECC effort.
A very brief account of the key issues pertaining to governing performance of
wind generation is presented here.
Conventional synchronous generators (in hydro, steam, nuclear, gas and
combined-cycle power plants) inherently add inertia to the system. This is not
necessarily true of all wind turbine generators. Conventional, direct connected
induction generator based wind turbine generators will add some inertia to the
system. Doubly-fed and full-converter units do not unless specifically designed to
do so. This is a consequence of the control action of the power electronics. For
doubly-fed and full-converter units (Le. units connected to the system through a
fully-rated frequency converter) the converter is designed to regulate the power
output of the unit tightly and extremely quickly (milliseconds). Thus, the electrical
output of the unit is kept constant throughout a typical frequency disturbance
and there is no effective unit inertial response. There are proposed techniques,
however, for augmenting the converter controls to emulate inertial response [22].
It remains to be seen how these can be effectively implemented in the field.
Similarly, wind turbine generators do not contribute to primary frequency control.
One example where such functionality has been demonstrated is the Horns Rev
off-shore wind farm in Denmark [21]. In this wind farm, the manufacturer (Vestas)
demonstrated various control features one of which is a reserve capability. That
is, it is possible to operate the wind farm to maintain, for example, a 5% reserve
margin that may be called upon during a frequency decline. The wind farm can
maintain a 5% margin between the power it generates and the actual power
available in the present wind conditions. This of course means that the amount
of megawatt reserve varies with the wind profile and is not constant. It could be
required that the controls, when possible, keep a specific megawatt margin in
reserve (e.g. always keep 5 MW in reserve). In either case there are clear
commercial consequences as well since the amount of power kept in reserve is
not utilized. This tends to suggest that on large interconnected systems it may
make more economic sense to maintain reserves on more conventional and
traditionally responsive units such as hydro and thermal plants. Nevertheless, in
small and islanded systems where wind is likely to be a major portion of the
generation mix these issues of inertial response and primary frequency control
require careful consideration and further development of the suited control
strategies.

1-20
1.6 Effect of Load on Frequency Regulation
This report is focused on the primary frequency regulation provided by turbine-
governors. However, one other important contributing factor to primary
frequency regulation should not be forgotten. This is an inherent physical
property of the system often called load damping (D). It represents the amount
by which total system load varies as a function of frequency. This can be
understood by considering the typical types of loads on the system. For
example, the mechanical load (and thus electrical power consumed by) most
induction motor driven loads is a function of frequency, such as a fan. As the
system frequency droops, the speed of the motor drops and so does the amount
of mechanical power required to drive the load at the lower speed.
A typical number assumed in most simulations for this effect is between 1 to 1.5%
change in load for a 1% change in frequency. So for example, if we assume that
a system has a total load of 100 GW, and that the load D is 1.5% for a 1% change
in frequency (on a 60Hz system), then this translates to 2,500 MW 1Hz.
Theoretically, if on such a hypothetical system 2,500 MW of load were lost, and
no generation governing is present, then the frequency would drop by 1 Hz. This
of course is a very simplistic example to illustrate the point. To operate a system
without good generation governing is dangerous. Without proper and
adequate generation governing to limit the frequency decline (rise) following a
large loss of generation/import (load), the consequences could be dire - namely
leading to a blackout. See for example the discussion on the Italian blackout in
section 2.6.3.
One last comment is also pertinent, with respect to load. Due to the voltage
dependence of load, during a major loss of generation there may be a
significant depression in local system voltage. As a consequence, the reduction
in load due to voltage decline also has a secondary regulation impact on
frequency decline. See for example the discussion on the dependence of the
simulation results, based on load characteristics, in section 2.2.4.

1.7 The Influence of Economic, Market, and Reliability Criteria


Considerations on Primary Governing Frequency Response:
Beginning in the 1960s, the electric utility industry has experienced a technology
transition. Prior to that time, there were two primary methods of producing
electric power, hydro and steam turbine technology. Since that decade there
have been a number of new technologies to enter the power production arena.
These initially included nuclear and simple-cycle gas turbines based upon jet
engine technology. Later these technologies were followed by large scale gas
turbines and combined-cycle technologies. Recently there has been a
significant movement toward renewable technologies including wind and solar
power production. Now there are indications that distributed generation
technologies may be on the near horizon. All of these technologies differ in
many significant technical respects from the hydro and steam turbine
technologies that were used from the birth of the industry for more than a half
century. Even though the majority of the production of electricity is still from

1-21
hydro and steam turbines, these newer technologies are increasing their
penetration into the electric markets. This penetration of new technologies is
contributing to the reduction in governor response that is being observed.
In the traditional vertically integrated utility structure, the reduced governor
response problem is easily solved. All that must be done is for the system
operator to convince the executive in charge of the generation of the need to
maintain adequate and dispersed governor response to insure reliability. That
executive would then use line "command and control" authority to instruct those
working for him to take care of the problem. The line "command and control"
would then pass this instruction down the chain of command and solve the
problem by changing operating methods for installed generation and creating
new methods to manage technology for new generation.
The problem of reduced governor response generally cannot be solved in this
manner today, because the transition facing the industry has not been a
transition in new technology alone. It has also included transitions in the way the
industry is structured. This transition in structure began with the initial opening of
the system to Independent Power Producers (IPP) through PURPA, and it has
continued with the deregulation and restructuring of the industry that is taking
place worldwide. As a consequence, the traditional decision making structures
and lines of authority are being reconfigured. The common goal associated with
all of these efforts is to improve the economic efficiency of the electric power
systems through this restructuring process. Some have suggested that the
technology transition influenced the transition in structure, but that is not
important with respect to reliable system operation. The only issue that is
important is the development of new decision making processes that will
properly support good electric system design, installation and operation. Under
these new structures, the "command and control" lines of authority have been
replaced by market relationships.
In addition, the traditional backstop of NERC Reliability Criteria may no longer be
able to support the provision of reliable primary governing response, because of
changes in the relationships between primary governing response and reserves
that has occurred as a consequence of penetration of these new technologies.
1.7.1 Market Structure Influences
In many of the restructured markets, the lines of authority have been severed.
Markets have required the divestiture of generation from transmission or at the
least separation of the two functions within the corporate organization. In other
cases, the system operations function has been separated and moved to an
independent entity, the ISO or RT04. In many markets the system operator no
longer has line authority over the generator. As a consequence, the oversight
that existed in the old vertically integrated utilities is non-existent. This by itself is
not necessarily bad. When this change in responsibility and structure is coupled
with the needs of the industry to provide a service that is necessary for reliability,
has significant operating cost and is undefined in most markets, the incentives to
provide this service disappear. Not only do they disappear, they are replaced

4 In Europe the TSO.

1-22
by disincentives in the form of additional costs that hurt the competitive position
of those generators continuing to provide primary governing frequency response
through governor action.
A good example of how this restructuring of the line of authority associated with
traditional command and control is highlighted by the EPRI Report on Ancillary
Services [23], [24]. This report demonstrates and quantifies the savings associated
with operating a steam turbine in sliding pressure mode and as a consequence
disabling the turbine-governor. Since these savings can be realized by changing
to this mode of operation by the independent generator, market
competitiveness creates incentives to operate in this manner reducing governor
response. Unless the market includes financial incentives to operate in a mode
that insures governing frequency response, measures the response provided,
and compensates for the provision of that response, there is a temptation to do
otherwise even when the rules prohibit such operation.
In the new market structures, with the line authority command and control
disrupted, the incentives to provide the necessary primary governing frequency
response for the system must be provided by market price signals. The current
assumption is that by solving an economic dispatch constrained by current
reliability rules, the economic incentives to operate reliably will be included in the
dispatch price. Unfortunately, the current reliability rules do not include
constraints on minimum primary governing frequency response requirements.
The first step in providing this incentive must be creation of primary governing
frequency response services that provide the economic communication
channel over which the incentive will be communicated. Simply, a market
without a contracted service for primary governing frequency response is
incapable of providing economic incentives that indicate how much of that
primary governing frequency response service is needed.
Markets have also taught us that infrequent or spot measurement of services
does not provide the optimum structure for the necessary incentives. This is
because infrequent or spot measurement provides an imperfect measurement
that passes the determination of whether or not the contracted service was
provided to a monitoring or enforcement function in the market. The
consequence is that whenever the monitor determines that the minimum limits
have not been met, there must be an enforcement decision with respect to how
to penalize the specific instance of non-compliance. This process adds
significant uncertainty whether a non-delivery penalty will be applied, and if it is
applied, uncertainty with respect to the magnitude of the penalty. This works
against fair enforcement, because it is well known that the most effective
deterrents are those deterrents that are enforced with certainty.
Since the inclusion of a primary governing frequency response service is required
to provide something upon which to attach the incentive, it is also important that
the method of measurement provide the greatest certainty possible with respect
to whether or not the service contracted was provided. This can be done in the
most effective manner using a continuous measurement method that correctly
measures the actual primary governing frequency response provided to meet
the contracted requirements.

1-23
1.7.2 Reliability Criteria Influences
Current reliability criteria are based upon a set of assumptions that the industry
has used for years that equate the ability to provide primary governing
frequency response with the provision of spinning reserves. This relationship was
reasonably consistent when all generating units were equipped with governors
with a 5% droop and were operated in a manner that assured the governor
would function. As the interconnections have expanded to include new
generation technology, and as operating methods have been modified to
improve economic efficiencies of individual generators, the relationship
between the holding of reserves and the ability to provide primary governing
frequency response has declined to the point where the provision of rules that
assure reserves cannot guarantee adequate primary governing frequency
response. Some examples of these changes are provided in this document and
in the NERC Frequency Response Standard Whitepaper [25]. Under these
circumstances, new methods that measure more than just reserves must be
developed to assure that adequate primary governing frequency response is
being held on the interconnections.
Traditional reliability criteria in North America required a specific amount of
operating reserve (reserve capable of being loaded in 10 minutes) 50% of which
is spinning reserve (synchronized reserve capable of being loaded in 10 minutes)
to insure adequate primary governing frequency response requirements for
reliability. In addition, in many regions, the amount of reserve that could be
provided by a single unit was also limited, for example to 20% of unit capability in
ERCOT. These criteria are a simplification of the true reliability requirements and
are based on certain assumptions with respect to the underlying technology
used to provide the reserve. The following examples demonstrate the effect of
these underlying assumptions.
1.7.2.1 1960's Example of Reliability Reserve Requirements
In the 1960's almost all of the generation was hydro or steam turbine technology.
When the reserve was provided by hydro all of the reserve capability was
available as spinning reserve and all of that capability was also available as
primary governing frequency response although that response could be limited
by the range of frequency operation on the interconnection. A 5% droop
provides a limit of about 20% of a generating unit's capability when the
maximum frequency change is limited to 0.6 Hz or 59.4 Hz on an interconnection
with a scheduled frequency of 60 Hz. As a consequence, in North America,
Spinning Reserves greater than 20% of a generating unit's capability are
effectively unavailable as primary governing frequency response. On the
Eastern Interconnection, where the frequency control is tighter, the available
primary governing frequency response is limited to significantly less than 20% of
the unit's capability. When the reserve is provided by steam turbine, not only is
the primary governing frequency response limited by the maximum frequency
change but it may also be limited by the amount of spinning reserve held on a
unit which is limited by the ramping capability of the unit. These ramping limits
were usually the result of limits in the steam plant rather than the turbine. A
steam unit with a ramping capability of 2% per minute would be limited to

1-24
providing 20% of the unit's capability as primary governing frequency response.
A steam unit with a ramping capability of 1% per minute would be limited to
providing 10% of the unit's capability as primary governing frequency response,
or it would only be allowed to account for providing reserve capable of
supporting primary governing frequency response equivalent to a 0.3 Hz change
in frequency. These effective limits are similar to the effective limits resulting from
maximum expected changes in interconnection frequency.
As a consequence of the correlation between the natural limits of the
technology and the reserve requirements, the specification of reserve was
equivalent to a specification of primary governing frequency response.
1.7.2.2 Example of Applying the Same Reliability Reserve Requirements Today
The provision of 10 minute reserves no longer assures sufficient primary governing
frequency response. A few examples of how the relationship between reserves
and primary governing frequency response has changed will clarify how the
traditional reliability reserve criteria fails to highlight this problem.
Some of the newer technologies allow much faster loading of generators than
was possible with the steam boiler technology of the 1960's. If for example a
generator can load at a rate of 5% of capability per minute, it would be able to
provide spinning reserve equal to 50% of the unit's capability while the 5% droop
on the governor still limits the primary governing frequency response to
something in the range of 10% to 20% of the unit's capability. Thus, it is possible
for only a fraction of the spinning reserve to be available as primary governing
frequency response today as compared to the amount available in the 1960's.
Both steam plants designed for cycling operations and gas turbines have
significantly greater loading rates than was available in the 1960's.
Operation in valves wide open sliding pressure mode eliminates all of the primary
governing frequency response from a unit, but still allows that unit to respond to
control signals at the natural response rate of the boiler which may be the
limiting characteristic of the unit for a 10 minute response.
1.7.2.3 Summary of Reliability Criteria Influences
As a consequence of new technology changing the relationship between
operating reserves and primary governing frequency response, current reliability
criteria based solely on reserve requirements provide a false sense of security
with respect to the reliability and security of the interconnections with respect to
primary governing frequency response. When these criteria are used in a
constrained economic dispatch, they fail to integrate the appropriate price
signals into the resulting energy price.
1.7.3 Ancillary Services Market Design
Ancillary Service markets have been designed to meet current reliability
requirements. As a consequence there are few markets worldwide that include
primary governing frequency response as an ancillary service. Although many
markets include discussions of responsive reserve (primary governing frequency
response) service, unless there is a measurement of that service in the market,
the service may not be provided. It takes measurement intervals less than one to

1-25
two minutes to effectively measure primary governing frequency response.
Therefore, most markets fail to include this necessary service simply because they
do not measure its delivery. Including a primary governing frequency response
service in the ancillary service markets will require new methods of specifying the
amount of service required to maintain reliability and new methods to measure
the confirmation of delivery of the service. In addition, since the primary
governing frequency response service can be substituted for other types of
reserve, it must be defined in a manner that enables the limited substitution of
primary governing frequency response for other reserve services when it is
economic to do so.
1.7.4 Measuring Primary Governing Response
Reasonable market management requires, that if a product or service is
purchased for delivery at a future time, the delivery of the product must be
confirmed by measuring whether or not the purchased product or service was
actually delivered at the time specified in the delivery contract. In addition, the
failure to deliver the agreed amount of service should result in some specified
automatic adjustment in compensation for the amount delivered. If this
measurement and adjustment cannot be performed consistently in the market,
the delivery of the product or service cannot be assured. This is probably the
most difficult characteristic of market design associated with the inclusion of
primary governing frequency response in a market design.
Recent work in the area of measuring primary governing frequency response is in
the literature. A series of papers that included the real-time estimation of primary
governing frequency response were published in 2001 by NIPSCo and Purdue
[26], [27], [28]. These papers indicated that the best periods for estimating
measurement intervals were in the range of one to two minutes. Since CPS1
data is captured at 1 minute intervals, it may be possible to effectively measure
primary governing frequency response using this CPS 1 data. On a similar note,
Energy Mark recommended that average primary governing frequency
response could be measured using one minute data and linear regression
techniques [29]. If primary governing frequency response is to be included in the
energy ancillary service markets, this work would need to be extended and
confirmed for applicability as part of that effort.
1.7.5 Summary of Economic, Market and Reliability Criteria Influences
The technologies used by the electric utility industry have changed significantly
since the 1960's. These changes in technology have changed the traditional
relationships between generation reserves and primary governing frequency
response, and have resulted in declining primary governing frequency response.
If current trends continue, reduction in primary governing frequency response
over time will eventually put all of the interconnections at reliability risk.
Structural changes in the electric utility industry have severed the traditional lines
of command and control between the system operator and the generation
provider limiting the ability of the system operator to address this decline in
primary governing frequency response.

1-26
Reliability requirements based on traditional relationships between primary
governing frequency response and reserves have also changed as the result of
newer technologies and operating methods reducing the ability of reserve
constraints to assure sufficient primary governing frequency response.
The most promising path to address this problem in the new market based
restructured environment is to include primary governing frequency response in
the restructured markets and the reliability criteria used to assure reliability in
those markets. Inclusion of the primary governing frequency response in these
markets will require the definition of new ancillary services and new
measurement methods to confirm the delivery of these new services.

1.8 Example of an Energy Market That Has Established Primary


Frequency as an Ancillary Service - New Zealand's Market for
Ancillary Services
New Zealand power industry is deregulated and the electricity market has been
in operation since 1996. Procurement and dispatch of ancillary services is
performed by the System Operator for ensuring the delivered power quality and
maintaining power system stability. The system operator purchases the following
ancillary services from ancillary service agents:
Frequency regulating reserves (secondary frequency)
Instantaneous reserves (primary frequency)
Over-frequency reserves
Voltage support
Black start
Primary frequency ancillary service is procured through a half hour clearing
market process, whereby the ancillary service agents submit offers to the system
operator. The market for that service is reconciled, priced and settled on a half-
hour basis for such quantities as the system operator assesses to be practicable
and cost-effective to procure. Before an offer can be submitted, the ancillary
service agent must enter into an ancillary service procurement contract for the
service. The ancillary service procurement contract will set out the offer, pricing
and settlement mechanisms for the ancillary service without stipulating specific
offer quantities.
The instantaneous reserves are computed and scheduled for every trading
period (30 minutes). In scheduling the reserves, the total cost of energy and
instantaneous reserves are co-optimized. (Note: As a result of co-optimization
process, there could be situations where some large units are partially
dispatched on order to reduce the cost of required reserves)5.

5 Reference [30] provides discussion and analysis related to reserve markets in general.

1-27
The actual procurement cost of instantaneous reserves for the period 1
November 2002 - 31 October 2003 was NZ$27 million.
1.8.1 New Zealand Power System
The small size of the power grid in New Zealand and its' generation mix requires
maintaining an optimum level of primary frequency ancillary service for efficient
and stable operation of the power system.
The electricity generation sources in New Zealand consist of a mix of plant types:
hydro, steam, gas, combined-cycle power plants (CCPP) and geothermal. The
total installed generation capacity is approximately 8,200 MW and Table 1-1
shows relative size of the installed generation mix. Maximum electrical load
supplied in 2006 was approximately 6,750 MW, of which 4,500 MW was in the
North Island and 2,250 MW was in the South Island.

Table 1-1. Generation Mix in New Zealand (2006)


Installed Capacity
Generation Type
MW %
Hydro 5,289 61
Steam (Stand-Alone Steam, Coal/Gas-Fired)) 1,304 15
Gas Turbines - Combined Cycle 903 10
Gas Turbines - Open Cycle 205 2
Geothermal 490 6
Wind 165 2
Co-Generation 354 4
Total 8,637 100

The largest generation units in the North Island include, two single shaft CCPPs
rated at 354 MW and 380 MW, and four steam turbine units each rated at 250
MW. Each of the CCPP units represents a significant proportion of the total North
Island generation at any given time, ranging from approximately 10% during
peak load periods to 20% during light load periods. South Island generation is
solely based on hydro plants with the maximum capacity unit rated at 120 MW.
The high voltage ac transmission networks in the two islands are connected
through a bipole HVDC transmission link. The HVDC power transmission is
predominantly from the hydro generation rich South Island to the industrial and
commercial load centers in the North Island. The maximum capacity of the
HVDC link presently stands at 1,040 MW.
1.8.2 Primary Frequency Ancillary Service
The following events are categorized as "contingent events" where the impact,
probability of occurrence, and the estimated costs and benefits of mitigation

1-28
are considered to justify implementation of processes (including procurement of
ancillary services) incorporated into scheduling and dispatch pre-event.
The loss of a transmission circuit
The loss of an HVDe pole
The loss of a single generating unit
The loss of both transmission circuits of a double circuit transmission line,
where past experience shows a high level of likelihood of occurrence.
Instantaneous reserves are procured and dispatched by the System Operator for
achieving the following quality conditions and limits, following the occurrence of
a contingent event:
no asset exceeds its stated capability
frequency in either island does not drop below 48 Hz (4% drop)
frequency in either island is restored to within 50 Hz +/- 0.75 Hz (1 .5 %)
within 1 minute
Instantaneous reserves are commonly comprised of one or more of the following:
interruptible load, partly loaded spinning reserve, or tail water depressed reserve.
The system operator procures instantaneous reserve as:
Fast instantaneous reserve (FIR) and
Sustained instantaneous reserve (SIR)
FIR generally refers to the additional capacity (in MW) available within six
seconds after a contingent event and sustained for a period of at least 60
seconds. SIR generally refers to the average additional output (in MW) available
during the first 60 seconds after the contingent event and which is sustained for
at least 15 minutes after the event.
The frequency response of the New Zealand power system following a
contingency event is such that it could be best managed by procurement of the
above two types of the reserves.

1.9 Including Demand-Side Participation as a Resource for Primary


Governing Frequency Response
This report mainly addresses historic problems with primary governing frequency
response models used to estimate the effects of primary governing on
interconnection frequency. Since most of the historic response has been
provided by generators and still is, these models and the problems associated
with inaccuracies in model parameters are related to generator governor
models. Nothing in this report should be taken to imply that only generation
governors are appropriate resources for primary governing frequency response.
It is apparent from the structure and material contained in this document that
the demand-side of the industry has been given little consideration with respect
to how they could contribute to the assurance that adequate primary governing

1-29
frequency response is available. The only mention of using demand-side
resources for primary governing frequency response, in the USA, is the ERCOT
LaaR service. This service uses under-frequency relays to interrupt firm load when
large frequency excursions occur. This is an excellent way to supplement the
frequency response of the interconnection but it does not offer a replacement
to the continuous frequency control during both disturbance and disturbance
recovery that a generator governor provides. All other references to the
inclusion of primary governing frequency response in the ancillary services are
linked directly to generation governors.
There are many types of loads that, if coupled with effective advanced control
systems that would measure the interconnection frequency and adjust the load
continuously in response to that frequency, could contribute to primary
governing. These could be excellent sources of primary governing frequency
response. The real questions are:
1. Would demand-side participation contribute to more reliable systems?
2. What are the costs and benefits of having demand-side participation?
3. What types of load could provide primary governing frequency response?
4. What technical work needs to be completed to enable demand-side
participation?
5. How would demand-side participation affect interconnection modeling?
6. How would loads be rewarded for providing this service?
7. How would the provision of this service be measured?
These questions are beyond the scope of this report. They appear herein simply
to emphasize the fact that this important service could be supplied in part by the
demand side of the industry. This is work that should be included as part of future
studies on this subject. We also note that some work has started in this area, for
example the "grid friendly appliances" program from Pacific Northwest National
tobs-.

1.10 Summary
This chapter has provided an overview of the "primary governing problem" -
namely the recent recognized disparity between actual system primary
governing response to generation/load imbalance and the simulated response
with power system planning models. For the sake of completeness, a description
has been provided of the behavior of all the major types of generation. In
addition, a discussion has been provided on the influence of economic, market
and reliability criteria considerations on primary governing frequency response.
The following chapter provides detailed accounts of case studies and
experience from various utilities that exemplify this disparity. In the last chapter,
the Task Force provides the major conclusions of this work and recommendations
for future work.

6 For more information see http://gridwise.pnl.gov/technologies/transactive controls.stm.

1-30
2 TECHNICAL STUDIES AND RESULTS
2.1 Introduction
This chapter presents a set of case studies and detailed accounts of experience
of various utilities from around the world on the primary frequency governing
response performance and practices in the present day interconnected power
systems.

2.2 Primary Governing and Frequency Control in ERCOT


2.2.1 Introduction
In this subsection, a brief description is given of the Electricity Reliability Council of
Texas (ERCOT) power system, some present practices for unit primary governing
and frequency control, a recent approach to modeling primary governing for a
study will be described, and a comparison between the actual system
performance during an event and the simulation of the event. While some
aspects of primary governing and responsive reserve will be covered, the major
emphasis will be on primary governing. Note that other services such as
regulation service will not be covered in this document, although they do
contribute to frequency control.
2.2.2 Background
The Electric Reliability Council of Texas, Inc. (ERCOT) is one of 10 electric reliability
regions in North America [31]. ERCOT serves about 85% of the electrical load in
Texas. The summer, 2003 peak hourly demand in ERCOT was 59,996 MW. The
overall generation capacity is approximately 70,000 MW. Table 2-1 summarizes
generation capacity in ERCOT by type [32].
An important characteristic of ERCOT is that it is completely located in the state
of Texas, and there are no synchronous connections to other regions. There are
two back-to-back HVDC ties connecting ERCOT to other reliability regions in the
USA. The total capacity of the HVDC ties is 856 MW. Thus, ERCOT can rely on
only limited help from other regions should an event result in significant
frequency decline. Primary governing and frequency control are important for
an isolated system, even a system the size of ERCOT.

2- 1
Table 2-1. Summer 2003 Capacity by Type
Type % of Total
Nuclear 6
Hydro Less than 1
steam - coal 19
steam - gas/oil 38
Combustion turbine? 5
Combined Cycle 30
Other, including wind 2

2.2.2.1 Primary Governing and Frequency Control


Within ERCOT, there are two primary documents that address how the system
operates: the ERCOT Protocols, and the ERCOT Operating Guides. The Protocols
contain rules by which ERCOT implements defined market functions including
operation of the power grid. The Operating Guides supplement the Protocols
and describe the working relationship between the ERCOT Control Area
Authority and entities within the ERCOT system that interact with the ERCOT
Control Area Authority. Note that both the Protocols and the Operating Guides
are living documents and are changed as the need arises. The information
below describes practices that were current at the time of writing, but may not
be accurate in the future.
In ERCOT, primary governor response and responsive reserve service are treated
as separate items. However, they are linked in part because on-line machines
are required to have their governors in service. The term primary governor
response is not explicitly defined, but requirements for primary governor response
function are contained in the Protocols and Operating Guides. Responsive
reserves are reserves that are intended to help restore the frequency of the
transmission system within the first few minutes of an event that causes a
significant deviation from the standard frequency. Standard frequency is not
defined in the Protocols or Operating Guides, but is presumed to be 60.0 Hz.
Key ERCOT requirements for primary governor response are as follows: [33], [34]
In Service: ERCOT requires that whenever a generation resource is on-
line, its turbine-governor shall remain in service and be allowed to
respond to all changes in system frequency. ERCOT defines
generation resources as facilities that produce energy and that are
owned or operated by a generation entity. Further, generation entities
shall not reduce governor response on individual resources during
abnormal conditions without ERCOT's consent, unless equipment
damage is imminent.
Reporting: Any short-term inability of a generation resource to supply
governor response shall be immediately reported to ERCOT.

7 The terms combustion turbine and gas turbine are used interchangeably through out
this document. Both terms refer to thermal units that derive their power from a process of
com busting natural gas (or liquid petroleum gas) with air.

2-2
Droop: The Operating Guides state that every effort should be made
to maintain governors with a five percent (5%) droop characteristic.
Governor Testing: Governor performance tests should be conducted
at least every two years. Sample test procedures for mechanical-
hydraulic and electro-hydraulic governors are in the Operating
Guides.
Poor Governor Response: The Operating Guides list the following
elements that can contribute to poor governor response:
- Governor dead band
- Valve position limits
- Blocked governor operation
- Control mode
- Adjustable rates or limits
- Boiler/turbine coordinated control action
They also state that every attempt should be made to minimize the
effects of these elements on the governor operation. Note that
combustion turbines operating on their temperature limit are not listed
as leading to poor governor response.
Overall Response: The combined response of all generation resources
interconnected in ERCOT to a measurable event (the sudden change
in interconnection frequency to a value between 59.700 Hz and 59.900
Hz or between 60.100 Hz and 60.300 Hz, and a frequency change
greater than or equal to +/- 0.100 Hz) shall be at least 420 MW / 0.1 Hz.
There is no mechanism for ERCOT to know whether this minimum
response is available. The actual performance is calculated as a post
event check of response.
Thus, there is an expectation that any generator that is on-line will respond to
sudden frequency changes with governor action. It is also recognized that there
will be exceptions to this expectation. These exceptions include generators
operating at their maximum or minimum power limit, units with impaired
governors, nuclear units, and wind units that have no governor. Experience has
shown that governors on set point control may provide an initial response to a
sudden change in frequency, but then return to the set point in 30 to 60 seconds.
Presumably, that response does not fulfill the ERCOT requirements.
Key ERCOT requirements for responsive reserve service (RRS) are as follows [33],
[34]:
In Service: Generation resources providing RRS must have their
governors in service.
RRS Obligation: The ERCOT responsive reserve obligation is 2300 MW.
Sources: Responsive reserve service may be provided by: (a)
unloaded generation resources that are on-line, (b) resources

2-3
controlled by high-set under-frequency relays, (c) hydro responsive
reserves, or (d) from dc tie response that stops frequency decay.
Generation Resources: Generation resources providing RRS must be
on-line and capable of ramping to the awarded output level within
ten (10) minutes of the notice to deploy energy, must be immediately
responsive to system frequency, and must be able to maintain the
scheduled level for the period of service commitment. In no case shall
more than 20% of the net dependable capability for any thermal unit
be used as responsive reserve. In no case shall more than 20% of the
net dependable capability for any hydro unit with a 5% droop setting
operating as a generator be used as responsive reserve.
Load acting as a Resource: Load acting as a resource must be
loaded and capable of unloading the scheduled amount of RRS
within ten (1 0) minutes in response to two initiating events: instruction
by ERCOT or by action of under-frequency relays. The under-
frequency relays are set to initiate the interruption whenever system
frequency reaches a specific value. The initiation setting of the relay
shall not be any lower than 59.7 Hz. Load interruption will occur within
20 cycles of the time that frequency decays to a value low enough to
initiate action of the under-frequency relay. In no case may
interrupted load be restored to service without the approval of the
ERCOT operator. The amount of resources on high-set under-frequency
relays providing RRS will be limited to fifty percent (50 %) of the total
ERCOT RRS requirement.
Deployment of RRS: Responsive reserve energy shall be deployed as
necessary to meet NERC requirements. This shall be accomplished by:
(a) automatic generation action as a result of a significant frequency
deviation; (b) through use of an automatic signal and a dispatch
instruction to deploy responsive reserve energy from generation
resources; and/or (c) by dispatch instructions for deployment of
responsive reserve energy from load acting as a resource via an
electronic messaging system to providers. Once RRS is deployed, the
obligation to deliver energy shall remain until specifically instructed by
ERCOT to stop providing energy from RRS, but not longer than the
period the service is scheduled. The HVDC tie-line response must be
fully deployed within fifteen (15) seconds on the ERCOT System after
the under-frequency event.
In addition, ERCOT has an under-frequency load shedding (UFLS) scheme to
provide load relief should unusual circumstances result in load exceeding
generation and normal corrective action, such as governor response and
responsive reserve service, has not been adequate to correct the problem.
Some key requirements for the UFLS scheme are as follows:
At least 25% of the ERCOT System Load that is not equipped with high-
set under-frequency relays shall be equipped at all times with

2-4
provisions for automatic under-frequency load shedding. The under-
frequency relays shall be set to provide load relief as follows:

Frequency Threshold Load Relief

5% of the ERCOT System Load


59.3 Hz
(Total 5%)
An additional 10% of the ERCOT System Load
58.9 Hz
(Total 15%)
An additional 10% of the ERCOT System Load
58.5 Hz
(Total 25%)

Load equipped with under-frequency relays should be dispersed


geographically throughout the ERCOT System to minimize the impact
of load shedding within a given geographical area.
Under-frequency relays connected to each load will operate with a
fixed time delay of no more than 30 cycles. Total time from the time
when frequency first reaches one of the values specified above to the
time load is interrupted should be no more than 40 cycles, including all
relay and breaker operating times.
Automatic Load restoration for a UFLS operation is not currently utilized
in ERCOT.
2.2.3 An Approach to Modeling Primary Governing
As indicated above, 50% of the 2300 MW responsive reserve requirement can be
provided by load acting as a resource. This 50% allowance was established as
the result of a request in 2002 for the ERCOT Dynamics Working Group (DWG) to
explore what percentage of the 2300 MW responsive reserve requirement could
be provided by shedding load at frequencies of 59.7 Hz and above. As with any
study, many decisions must be made on how to handle various issues related to
the study. The following describes the conditions at the time of the study, an
approach used to model primary governing for this study, and why the
approach was used.
The previous ERCOT spinning reserve study (1988) used governor data as
reported from generation owners and contained in the ERCOT dynamics data
file. Governor action was blocked for some units known to have little or no
response during system frequency disturbances. At that time, deregulation had
not occurred, most generation was from conventional steam boiler driven
generators, and integrated utilities owned most of the generators. The models
used to make the study assumed steam pressure would be 100% under all
conditions for the governor to control [35]. That assumption can be
approximately correct for conventional boilers operated in "boiler follow" control
mode.
Several things had changed in the 15 years since the previous study.
Deregulation in ERCOT had occurred. Large amounts of generation was now

2-5
owned and operated by independent power producers. Combined-cycle
plants provided a significant part of the total generation in ERCOT. Generation
from wind plants was increasing steadily. These changes prompted an
evaluation of how the generator governors should be treated in the study.
Observations of the Eastern Interconnection had shown a decrease in system
response to frequency over a five year period [36]. One utility determined that
their response was 30% of expected response [1]. Similarly, the ERCOT response
was not as expected [37]. This decline in governing response has been
attributed to a quest for efficiency [36]. This quest for efficiency can manifest
itself in such ways as changing unit operating modes from "boiler follow" to
"sliding pressure", operating at 100% output, and in the case of combustion
turbines, operating on temperature control [37], [4]. While this decline in response
was not a concern for the large Eastern Interconnection (527 GW in 2001), it is a
concern for a smaller system such as ERCOT (53.4 GW in 2001). This prompted
discussions with operators from three "utilities" representing well over 60% of the
installed capacity in ERCOT. The responses indicated that "most" to "almost all"
conventional steam plant operating modes have been changed to a less
responsive mode (turbine follow or sliding pressure, for example). Experience
confirms these qualitative assessments [37]. It was determined that the nuclear
plants are operated such that they will not respond to system frequency
changes. Wind generation output is dependent upon the wind velocity, a
quantity that is virtually uncontrollable. Thus, wind units are not expected to be
responsive to frequency deviations or contribute to frequency restoration.
Combustion turbines and combined-cycle plants constitute a significantly larger
percentage of the installed capacity in ERCOT compared to 1988. A review of
the models and data available for this equipment, and literature concerning this
equipment resulted in several concerns. Combustion turbine output is
dependent upon several factors. One of these factors is the power system
frequency. The maximum megawatt capacity of a combustion turbine rises
when the frequency increases, and decreases when the frequency decreases
[20], [38]. An incident such as loss of a large generator or plant can cause a
significant frequency decrease. A combustion turbine providing spinning reserve
could have an output less than expected because of this frequency
dependency. Thus, at the critical time when more power output is needed to
arrest the frequency decline, it may not be available from the combustion
turbine. After a loss of generation incident, there is the possibility of frequency
overshoot (frequency greater than 60 Hz), especially when load shedding is part
of the frequency recovery plan. Without proper control design, the combustion
turbine could experience "flame out", and trip off-line" [1]. While these
characteristics of combustion turbines do not preclude them from providing
responsive reserve service, it was clear that the frequency effects needed to be
accounted for in this study. Unfortunately, no combustion turbine-governor
model available to the DWG models these combustion and frequency effects.

8 Email from M. Connolly to ERCOT Dynamics Working Group, February 19, 2002.

2-6
Combustion turbines also have exhaust temperature controls which can override
the governor controls [1], [38]. Several combustion turbine models available to
DWG model the exhaust temperature controls. Judging from simulations using
owner submitted data, most combustion turbines in ERCOT will be operated on
their exhaust temperature limit. This situation is to be expected since many
combustion turbines are part of combined-cycle plants. Operating at the
exhaust temperature limit usually increases efficiency. It is possible to operate a
combustion turbine at an exhaust temperature lower than its limit [38]. However,
the turbine could reach its exhaust temperature limit while responding to
frequency drop, thus affecting its response. For study purposes, the exhaust
temperature controls pose a problem similar to that of conventional steam
turbines: identifying and quantifying the actual controls in use.
ERCOT also has a significant amount of generation installed as combined-cycle
plants, both multi-shaft and single-shaft designs (refer to section 1.5.4 for a
description of combined-cycle power plants). Adequate single-shaft models
were not available to the DWG. For both types of combined-cycle units, the
steam turbine driven generator output is dependent upon the heat from the
combustion turbine. Usually, the steam turbine is operated in "sliding pressure"
control (see section 1.5.4). Because of the large heat recovery steam generator
time constant, the steam turbine will likely not respond in the time frames
appropriate for this study. For multi-shaft units, separate, distinct models are
usually used for the combustion turbine-governor and steam turbine-governor.
From a modeling perspective, simply removing the steam turbine-governor
model from the dynamics data will prevent the governor from responding.
Given the conditions mentioned above, it was clear to the DWG that the
approach to governor modeling for this study needed careful consideration.
Three alternatives were considered. First, an approach similar to that used in the
1988 study could be used. That approach generally used the governor models
and data as they existed in the dynamics data base. Governor action was
blocked for some units known to have little or no response during system
frequency disturbances. For the study, no comprehensive list of governors to be
blocked was available. In addition, simply using the governor models and data
as they exist in the dynamics data base would result in the simulated
performance of conventional steam generators being substantially better than
the actual performance for units where the operating point or control mode has
chonqed". Similarly, the lack of models which include combustion and speed
effects for combustion turbines would result in simulated performance being
better than actual performance. Decisions based on overly optimistic
performance could result in larger frequency swings and larger loss of load
should a significant loss of generation event actually occur in ERCOT. Given how
important governor and responsive reserve is to the reliability of the ERCOT
system, this option did not seem desirable.
Another option considered was to have all conventional steam plants tested in
the mode they usually operate. From this testing new models could be chosen

9 Email from M. Connolly to ERCOT Dynamics Working Group, February 19, 2002.

2-7
which more accurately represent their performance. However, testing would
cost thousands of dollars per unit and take many months to complete. Some
owners might not be willing to participate. and testing conventional steam units
would do nothing to correct the deficienc y in the combustion turbine-governor
models. Efforts to obtain more accurate combustion turbine models have been
underway for several years. Progress was being made. However, it was
considered unlikely such models would be available for the study, and indeed
that was the case. So, while testing and model development would likely have
produced the most accurate data and minimized the number of necessary
assumptions, this option was considered infeasible.
The third option considered involves a significantly different approach to the
study. The first two approaches essentially assume the models and related data
accurately represent each individual unit. Different units have different response
characteristics. Some units respond quickly, some slowly. Some units will pick up
large amounts of load, while others will pick up little load. Simulations essentially
"average" the response. An inherent problem with that approach is one does
not know in advance which type of unit will be providing responsive reserve. It is
possible that virtually all of the connected generation has poor response
characteristics 10. The approach used for this study was to define the minimum
response for units providing responsive reserve. Advantages of this approach
include:
1. It is not necessary to know in advance which units are providing
responsive reserve, or to be overly conservative by modeling only
"poorly" responding units.
2. It is not necessary to know the operating mode of the conventional
steam turbines.
3. Highly accurate models and data for each unit are not required. Note
that this statement does not apply to other types of studies.
4. It is not necessary to resolve the deficiencies in the available
combustion turbine models.
5. Assuming there are no other failures, the minimum system response to
loss of generation will be well defined and understood.
6. The minimum response characteristic is known. Units that do not meet
the minimum response can, within physical limits, be changed to meet
the minimum response requirements if the owners so choose.
7. Because the only important parameter is the response characteristic,
the specific model and data used for simulations can be treated as a
"black box". (For a given input to the black box, the output is known.
Specific knowledge of the circuit or system in the box is not necessary.)
Defining the minimum response essentially involved:
Removing all governor models from the ERCOT dynamics database.

10 Email from M. Connolly to ERCOT Dynamics Working Group, February 19, 2002.

2-8
Choosing one governor model to be used with all units allowed to
respond to frequency changes.
Tuning the governor model to have the desired response.
Selecting a broad range of units to have governor response
Adding the selected, tuned governor model to the dynamics
database for the responding units.
Making minor changes to the load flow case so 1150 MW to 2300 MW
of machine response was available.
For this study the IEESGO model was chosen to be the single governor model.
This is a fairly straight forward model originally developed for conventional steam
turbines. An advantage of this model is sufficient, relatively simple parameters
making the tuning process fairly easy. It also happens to be the most commonly
used governor model in the ERCOT system.
Choosing the response characteristics for the model was an important decision.
If the response characteristic is too fast, some units may not be able to meet the
minimum response requirements. If it is too slow, the system and customers could
be unnecessarily affected should a significant loss of generation occur. Several
approaches were considered. The approach used was to tune the model
parameters to match plots of actual loss of generation incidents. Specifically,
the February 8, 2000 loss of Oklaunion unit 1 and the April 8, 2000 loss of Martin
Lake SES unit 2 were used. With this approach, the minimum response
characteristic is essentially the average of the individual responses at the time of
these loss of generation incidents. In effect, this method defined an average
rate of response for units providing responsive reserve (6.25% of unit proven
capability per second).
The ERCOT Reliability and Operations Subcommittee (ROS) formulated
recommendations based on the DWG study. The recommendations led to the
current allowance that up to 50% of responsive reserve can be load acting as a
resource. ROS did not recommend a performance criterion for units providing
responsive reserve.
2.2.4 Comparing Actual and Simulated Response
On May 15, 2003 at about 2:52 AM, an electrical fault occurred on the 345 kV
line between Comanche Peak Switchyard and Parker Switching Station 11. Most
faults are cleared promptly with little effect on the power system. Unfortunately,
on May 15, relay failure lead to a series of events that resulted in the loss of about
7,200 MW of generation capacity and a system frequency decline to 59.26 Hz.
Since the frequency dropped below the first under-frequency load shedding set
point, about 1,549 MW of firm load was shed'". This May 15th incident may have
been the first time the under-frequency load shedding scheme operated in
ERCOT.

II Email from M. Connolly to ERCOT Dynamics Working Group, February 19, 2002.
12 Presentation by S. Francis to the Fort Worth IEEE PES at CPSES, July 8, 2003.

2-9
In the early morning of May 15, 2003, there were thunder storms and lightning in
the area. The ERCOT load was about 33,300 MW. The incident started when
apparently, a Blue Heron (a large bird with a 60 to 84 inch wing span) made
contact with the B phase (and shield wire) of a 345 kV line at a tower about 4
miles from Comanche Peak Switchyard. The initial single phase fault evolved
into a line to line to ground fault and eventually a three phase to ground fault.
Details may be found in [39].
The ERCOT dynamics database is a collection of well tuned models and data.
The models and data have been provided by the equipment owners, usually
based on manufacturer's typical data. The database contains tested values,
such as governor droop, machine maximum real power output, and machine
reactive output for some machines. Some common pitfalls such as missing
models and data, application program default parameters, or governor Pmax
value greatly exceeding the machine's actual capability are almost completely
absent.
Each simulation of the incident described below included the detailed
sequence of events from reference [39]. The load response to frequency
change was included in the simulations. At the time of the incident, models for
load acting as a resource or under-frequency load shed relay models were not
normally included in the ERCOT dynamics database. Since both these operated
during the event, models for both were added to the database. The reported
amount of load acting as a resource that tripped was 683 MW [40]. The models
added to the dynamics database tripped that amount of load 20 cycles after
the frequency reached 59.7 Hz. The under-frequency load shedding models
were built so that load would be tripped at the frequencies, in the amounts, and
with the time delays as indicated in the ERCOT documents discussed above.
For the initial simulation, the ERCOT dynamics database was used with the
additions described above but without any modifications to the data. The
network equivalent fault admittance was calculated for line to ground and line
to line to ground faults at the fault location, but fault impedance was assumed
to be zero. The results of the initial simulation are shown in Figure 2-1 .
The upper curve in Figure 2-1 , called "sim 2", shows the simulated frequency
response for the event. The lower curve, called "recorded", is the actual
recorded frequency response. The important frequencies of 59.7 Hz and 59.3 Hz
are high lighted as well. Obviously, the simulation frequency response is not a
good match to the actual frequency response. The relatively large initial
frequency increase in the simulation suggests the fault impedance was
significant and should be included in the simulation.
Figure 2-2 shows the same information as Figure 2-1 with the addition of a
frequency response curve, called "sim 21", where the simulation includes fault
impedance. As can be seen the initial frequency rise in the sim 21 curve is much
closer to the recorded value, and the overall shape of the curve is closer to the
recorded frequency response. Note that the simulated maximum frequency
drop is about half the actual frequency drop. This is significant because the

2 - 10
simulation suggests only load acting as a resource w ill be shed as a result of this
event .
No Data Changes

603

602
~ recorded

. sim 2
60.1

60

59.9

59.8
N
:I:
59.7

59.6

594

59.3
I
592 .......!
0 10 15 20 25
Seconds

Figure 2-1. Comparison of Recorded (blue) and Simulated (magenta) response to the
May 15, 2003 Event in ERCOT. The simulated response here is based on the original
ERCOT database with the addition of UFLS and load acting as a resource modeled.
Fault Y Adjusted

................................
!
602
1=;~~O:edl
Sim 21
60 1

60

599

598
N
:I:
59.7

596

595

59 4

59.3

592
0 10 15 20 25
Seconds

Figure 2-2. Comparison of Recorded (blue) and Simulated


(magenta and yellow) Response to the May 15, 2003 Event in ERCOT.
The simulated response here is based on the original ERCOT
database with the addition of UFLS and load acting as a resource
modeled. The difference between the magenta and
yellow curves is that for the yellow curve the
fault impedance of the initiating fault has been modeled.

2 - 11
During the actual event, a significant amount of firm load was shed via under-
frequency load shedding relays. Clearly, the simulation gives overly optimistic
results. A likely reason for this optimistic simulation result is the dynamics data,
specifically the governor droop value. While there are some minor variations, the
droop setting for most governor models in the ERCOT dynamics database is
about 5%. The reported system frequency regulation overall droop for this event
was 17% [40] .
The response during this event of 28 units owned by 6 different entities, all
connected to the same transmission provider, was examined . The qualitative
results revealed the following :
12 units had a small initial MW inc rea se response to the event, and
then returned to their pre-fault MW output within 2 to 30 seconds after
the event start.
9 units did not respond .
4 units responded quickly
3 un its responded with a small MW increase or responded slowly
The qualitative survey clearly suggests the reported droop number of 17% is
closer to the actual performance than the 5% in the database. To test the
hypothesis that the droop value in the dynamics database is to o optimistic, the
droop value was changed from 5% to 17% for every governor in the database.
The results of the simulation with 17%droop are shown in Figure 2-3.
Fault Y Adju sted, Droo p is 17%

59.9

59.8

59.7

:l!
59.6

59.5

59.4

59.3

592
0 10 15 20 25
Seconds

Figure 2-3. Comparison of Recorded (blue) and Simulated (magenta and yellow)
Response to the May 15, 2003 Event in ERCOT. The magenta curve represents simulated
response with the original database value of 5% droop on most units. The yellow curve
represents simulated response based on assuming a droop of 17% on all responsive units.

2 - 12
As in Figure 2-2, the recorded curve and the sim 21 curve show the actual
recorded frequency response and simulation with 5% droop and fault
impedance included, respectively. Curve sim 23 is the simulation results with fault
impedance included and a droop value of 17% for all governors.
The sim 23 curve in Figure 2-3 is much closer to the recorded frequency response
than the previous plot. However, the simulated frequency still does not reach
the first load shed step. Because some combustion turbines were on their
temperature limit, and a few conventional units did not have a governor model
in the simulation, changing every governor to 17%droop likely results in an overall
droop greater than 17%. Thus, a further increase in the droop value would not
be realistic or consistent with the overall measured value. There are numerous
governor model parameters that could be adjusted in an attempt to get better
fidelity between the simulation and recorded data. The recorded data is
primarily system wide data, so adjusting data on a unit basis beyond droop is
probably not appropriate.
There is one other system wide parameter that can be realistically adjusted, at
least crudely, that could affect the simulation results. That is the load model.
The "standard" ERCOT load model used for dynamic simulations considers all real
loads to be some combination of constant current and constant impedance.
Obviously some load in ERCOT is constant MVA. There is no known data that
indicates the percent of constant MVA load in ERCOT in the early morning in the
spring . Based on the latest load characterization for summer and winter, a
constant MVA load of 50% is a reasonable estimate. The results of a simulation
with the load model including 50% constant MVA load is shown in Figure 2-4.
Fau lt Y Adj usted, Droop is 17%, MVA Lo ad

60 1 - --------------------------------------------------l

60 -+- recor ded


. sim 23
sim 24
599

598

59.7
N
:I:

596

595

594

593

59.2
0 '0 15 20 25
Seconds

Figure 2-4. Comparison of Recorded (blue) and Simulated (magenta and yellow)
Response to the May 15, 2003 Event in ERCOT. The magenta curve represents simulated
response based on assuming a droop of 17% on all responsive units . The yellow curve is
simulated response based on assuming a droop of 17% on all responsive units and
assuming 50% ofthe system load is constant MVA.

2 - 13
As in previous figures, the "recorded" curve is the actual recorded frequency
response. The "sim 23" curve is the simulation results with fault impedance
included and all governor droop values set at 17%. The curve "sim 24" is the
simulation results with the same features as "sim 23" plus 50% of the load
considered to be constant MV A. The "sim 24" curve is a fairly close match to the
"rec ord e d " curve. Additional adjustment of model data would likely yield an
even closer match. Note that "sim 24" briefly drops below 59.3 Hz, the value
where under-frequency load shedding begins. No load was shed at 59.3 Hz in
the simulation because frequency recovered above the set point in less than 30
cycles.
From the above it is clear that when simulating events that include faults of long
duration, including the fault impedance in the simulation can be important.
Using a general system-wide approach, changes in governor droop and load
model were made, and each change improved the fidelity of the simulation
with the actual recorded frequency data. This suggest that governor model
data that represents the actual performance of the machine and load model
data that represents the actual load characteristics is important if simulations are
to be reasonably close to replicating the system response to actual events.
Clearly for this event, performing simulations using the "standard" ERCOT
dynamics data and load model provide results that significantly underestimate
the severity of the event.
Recorded , Ini tial, Adjusted

60.3

60.2

60.1

59.9

59.8
N
:I:
597

59.6

59.5

594

59.3

592
0 10 15 20 25
Second s

Figure 2-5. Graphical Summary of this Section. The blue curve is the actual system
response to the May 15, 2003 event. The top (magenta) curve is the result if one were to
simulate the event using the "standard" EReOT models and data. The "adjusted" curve
(yellow) is the result of a simulation using all the data adjustments mentioned i.e.
assuming 17% droop on all units and 50% constant M VA load throughout the system.

Figure 2-5 graphical summarizes th is section. The top curve would be the result if
one were to simulate this event for some planning study using the "standard"

2 - 14
ERCOT models and data. The "adjusted" curve is the result of a simulation using
the data adjustments mentioned above, and the "recorded" curve is the actual
frequency response to the event.
2.2.5 Summary
A brief description of ERCOT has been given to orient the reader. One important
feature of ERCOT is that it has no synchronous ties to other reliability regions.
Thus, primary governing and frequency control are important for ERCOT.
A listing of key requirements for primary governor response has been given to
explain how this important subject is addressed in ERCOT. From these
requirements, it can be seen that there is an expectation that any generator that
is on-line will respond to sudden frequency changes with governor action. It is
also recognized that there will be exceptions to this expectation. These
exceptions include generators operating at their maximum or minimum power
limit, units with impaired governors, nuclear units, and wind units that have no
governor.
A listing of key requirements for responsive reserve service response has been
given to explain how this important subject is addressed in ERCOT. In addition,
the ERCOT under-frequency load shedding program was described. While not
intended to be used except as a last resort to arrest system frequency decline, it
is an important part of the frequency control strategy in ERCOT.
Next, an approach to modeling primary governing from an actual ERCOT study
was presented. That section described the conditions at the time of the study,
an approach used to model primary governing for this study, and why the
approach was used. The approach involved removing all governor models from
the database and adding a governor model tuned to have a response typical
of the average total regulation response in ERCOT. In effect, this method
defined an average rate of response for units providing responsive reserve. This
approach removed many uncertainties associated with the more traditional
approach of using governor data as submitted to ERCOT and unmodified.
The frequency response of an actual event in ERCOT was compared to a
simulation of the event. It was found that the simulation did not have good
fidelity with the actual system response. Adjustments to the governor model
data, load model data, and fault impedance were made to improve the
accuracy of the simulation. Based on the success of these changes, it was
concluded that when simulating events that include faults of long duration,
including the fault impedance in the simulation can be important. Using
governor model data that represents the actual performance of the machines,
and load model data that represents the actual load characteristics is important
if simulations are to be reasonably close to replicating the system response to
actual events. Clearly for the event described, performing simulations using the
"standard" ERCOT dynamics data and load model provide results that
significantly underestimate the severity of the event.
Finally, in sections 2.2.3 and 2.2.4 two methods were presented to obtain
simulation results that would be closer to the actual system response than using

2 - 15
the existing governor models and data. These steps were taken because of
clear indications in some cases the available governor models and data were
inadequate or incorrect, and specific knowledge of the operating mode of
many machines was not available. This, however, does not mean that these
methods are recommended for general use. Chapter 3 contains
recommendations addressing the quality of governor data and models, and
additional information, such as machine operating mode, considered necessary
to perform simulations with reasonable confidence that they will match actual
system performance. Because the recommendations in Chapter 3 are not yet
fully implemented in all regions, the governor model and data changes
mentioned in sections 2.2.3 and 2.2.4 are included to illustrate that carefully
devised "stop-gap" measures can significantly improve the fidelity of the
simulations in some cases.

2.3 Nordel Interconnection


The Nordel system comprises the power systems in Denmark, Finland, Iceland,
Norway and Sweden. The synchronous system has an extension of about 2,000
kilometers in the north-south direction and 1,500 km in east-west direction . The
number of inhabitants amounts to about 20 millions.
There are no interconnections between Iceland and the rest of the Nordel
system . The power system in Denmark consists of one system in the eastern part
Zealand, wi th AC interconnections to Sweden and a HVDC link to Germany, and
one system in the western part Jutland, with AC interconnections to Germany
and HVDC links to Norway and Sweden.
2.3.1 The Power System
Figure 2-6 shows the installed capacity in the Nordel system at the end of year
2000. The total installed capacity in the Nordel system was then equal to 88.5
GW .
Installed Capacity 20001231

40 . - - - - - - - - - - - - - -- - - - - - - - - - - - - - ,

Total Install ed Cap acity =88.5 GW 30 .9


~

-
Q. 30
>-
'0
III
~ 20 16.6
o
"0
ell

~
!II
10
c
1.4

Denmark Finland Iceland Norway Sweden

Figure 2-6 . Total Installed Capacity at the End of Year 2000

2 - 16
The insta lle d capacity in Norway and Sweden amounts to about 30% and 35%
respectively of the total installed capacity. The sum of the installed capacity in
Finland and eastern Denmark amounts to about 25% o f the total installed
capacity in the Nordel system .
Figure 2-7 shows a breakdown of the installed capacity in the Nordel system. The
installed capacity of hydropower amounts to about 55% of the total installed
capacity in the Nordel system while the installed capacity of nuclear power
amounts to about 15%. The installed capacity of various forms of fossil fired
thermal power amounts to about 30% of the total installed capacity.

Installed Capacity 2000-12-31

60 -, -- - - - - - - - - - - - - - - - - - - - - - ---,

~ 50 _ 47 .7 Tota l Installed Capacity = 88.5 GW


>.
~ 40 -
u
III
g- 30
U
~ 20 -
13.8
III
iii 10 - 5.4 4 .1
s:::: 2.7 2.8
o ..
.. Ql
.
til .. Cl
c ~ .g
-
.sCl i::' III
Ql
-o c..U;
"t:l ~ Ql Ql
>-0 ~ "iii ~ J:--
III C
J: C. 0 c Ql
u .~ :ll J: ::J
l!l
til .-

~ c. ~
.a
Ql "t:l,c U"t:l
"t:l 0 C
C C.
o
U

Figure 2-7. Type of the Installed Capacity at the End of Year 2000

2.3.2 Load Frequency Control in Nordel


There is no AGC system in the Nordel power supply system [41]. The load
frequency control is based on : (1) primary control in power plants and (2)
manual secondary control based on the setting error [MW]:
SE=M+R*AJ (1)

Here R [MW 1Hz] is the frequency response characteristic in the country (area) .
The setting error for a country (area) indicates the difference between the sum
of all power settings on the generating units and the sum of total load plus losses
in the country (area) . The setting error (SE) used in the Nordel system
corresponds to the area control error (ACE) used in North America and
elsewhere. The difference is that the frequency response characteristic in
equation (1) is positive (R>O) and that it is expressed in MW 1Hz instead of
MW 1(0.1 HZ) . The natural frequency response of the load in the area is not
included in the frequency response characteristic R in equation (1).

2 - 17
The manual secondary control is based on the setting error as given in equation
(1). Figure 2-8 illustrates the control characteristics for a country (area) in the
Nordel system. The graph is similar to the graphs Cohn used to illustrate the
evolution of real-time control characteristics, see Figure 1 in reference [42]. The
setting error is equal to zero on a straight line that passes through the point (Pe.fe}.
where Po is the sum of the scheduled power transfer on the tie line out from the
area and f o is the scheduled system frequency.
The basic idea is to reduce the number of manual control actions and rely on
the automatic primary control action as long as the setting error, its integral and
the time deviation remain within certain limits. There is an individual deadband
for each country (area) and the country (area) is not required to act as long as
its setting error is within this deadband. A country (area) must not lower its
generation if such an action would cause the system frequency to drop below
the scheduled frequency. A country (area) must not rise its generation if such an
action would cause the system frequency to rise above the scheduled
frequency.
Norway and Sweden have the main responsibility to act if the system frequency
deviates too much from its scheduled value. The same holds true if the integral
of the frequency (time error) exceeds certain limits. A typical control action is to
start and synchronize a hydropower unit and adjust the gate opening so that the
unit operates close to the point of best efficiency.

Frequency

Setting Error = 0

Lower

Ra isa
I

'min --------------. ------------t----------------~-...........~---


I
I
I
I
: Sum of Power on Tie-Lines
I

Figure 2-8. Control Characteristic for the Areas in the Nordel system

The countries in the Nordel system have agreed that each area should
contribute to the primary frequency control by providing a minimum value of the
natural frequency response from the synchronized generating units according to
Table 2-2.

2 - 18
Table 2-2. Requirements on Frequency Response in the Nordel system

Country (Area) Frequency Response [MW/Hz]


Denmark (Zealand) 270
Finland 1,050
Norway 2,220
Sweden 2,460
Nordel 6,000

The ambition is that each country (area) shall contribute to the primary
frequency control in proportion to the energy consumption in the country (area).
The work reported here aims at finding a method to check the frequency
response characteristic of a generation unit during normal operation. The only
requirement is that the dispatcher refrains from changing the setpoint of the
output power during the measurements.
2.3.2.1 Recording from the 1983 Blackout in Sweden
On 27 December, 1983, a serious event occurred in the Swedish power system.
The total load in Sweden at that time was about 18,300 MW. A busbar fault near
Stockholm caused cascaded tripping of transmission lines in a critical region.
Protection systems automatically tripped tie lines to neighboring countries and
separated the Nordel system into several subsystems. The subsystem south of
region with cascaded line tripping had a power deficit of 7,000 MW. Frequency
and voltage dropped quickly. The under-frequency load shedding system did
not act as expected and no load shedding took place. All nuclear power units,
except one failed to trip to houseload operation. A blackout occurred in the
southern part of Sweden and the eastern part of Denmark. Kearsley [43] and
Wolve [44] give further information.
The frequency recording is shown in Figure 2-9 and corresponding voltage is
shown in Figure 2-10 [45]. Some time elapsed after the primary busbar fault
before the recorder started (the time 0.0 in the figure is the local trigger time).
The system probably separated at time 2.0 seconds, 2 seconds later the blackout
was a fact.

2 - 19
Frequency Recording in the South of Sweden
1983-12-27
51

'\
~ 50
>.
(J
c
Ql
;:,
lJ
...

\
Ql 49
U.

48
-1 o 2 3 4
Time [s]

Figure 2-9. Frequency Recordings from the 1983 Blackout in Sweden

The rate of change of network frequency increased gradually and was higher
than 3 Hz/s at time equal to 3 seconds. At the same time the rate of change of
voltage was about 200 kV Is (0.5 puis).

Voltage Recording in the South of Sweden


1983-12-27

425 , -- - ..- - - - -

I
400 -j-'
I
- - -, = :::::::::::;::-- - - - - -
!

~ 375 +1 ,~---- ----i

-
CI
I'll
"0
350 I!
> !
325 +-:- - -- - - -- - -- -- - -i \ - -- ----j
i
300 ..L.
! - - -- - - - - , - - - - -- - - - - ' - - '- - -----'

-1 o 2 3 4
Time [s]

Figure 2-10. Voltage Recordings from the 1983 Blackout in Sweden

2.3.2.2 Loss of a Single Generating Unit


Figure 2-11 shows a typical frequency response after losing the biggest
generating unit operating at about 1,050 MW in the Swedish system on 15
September, 1988.

2 - 20
Frequency Recordin gs in the South of Sweden
1988-09-15
50.2 T,.. T

N 50.0 .I-------:---=---+---+--~---+--------j

::s>.
o
49.8
::::l
C"
...
C1)

LL 49 .6

49 .4 ..I...---.;--- - - -- + ----i-- -----+---------'


o 2 3 4 5 6 7 8
Time [5]

Figure 2-11. Frequency Response after Loss of a Big Generating Unit

The rate of change of system frequency was only about 0.1 Hz/second and
much lower than the rate of change of frequency at the 1983 blackout in
Sweden.
2.3.2.3 Recordings from the 2003 Blackout in Sweden
At midday on Tuesday, September 23,2003, the Nordic power system
experienced the most severe disturbance in 20 years. The southern part of
Sweden and the eastern part of Denmark, including its capital city of
Copenhagen, were blacked-out. The cause was a close coincidence of severe
faults leading to a burden on the system far beyond the contingencies regarded
in normal system design and operating security standards. Very briefly, the
failure of a bus disconnector resulted in a double busbar fault on the 400 kV
system, which eventually culminated in the blackout (see [46] for more details).
The consequences to the power system from the disconnection of the busbars
were that the two nuclear units with a total output of 1,750 MW were tripped and
that the grid lost its transmission path along the west coast . Initially this triggered
heavy power oscillations in the system, very low voltages and a further drop in
frequency down to a level slightly over 49.00 Hz where under-frequency load-
shedding schemes start to operate.
The grid was then heavily overloaded on the remaining southeast and south-
central parts in terms of capability to sustain the voltages. This part of the grid
had no major generation connected and thus the reactive power support was
weak.
During some 90 seconds after the busbar fault the oscillations faded out and the
system seemed to stabilize. Meanwhile the demand in the area recovered
gradually from the initial reduction following the voltage drop by action of the
numerous feeder transformer tap-changers. This lowered the voltage further on

2 - 21
the 400-kV grid down to critical levels. Finally the situation developed into a
voltage collapse in a section of the grid southwest of the area around the
capital city of stockholm.
Within seconds following the voltage collapse, circuit breakers in the entire
southern grid were tripped from distance protections and zero-voltage
automatic controls. The interconnection to Zealand was disconnected as well.
This system was heavily affected by the transient conditions on the Swedish grid
and it did not manage to island itself to a stable situation before it broke down
completely.
Basically all supplies south of a geographical line between the cities of
Norrkoplnq in the east and Varberg in the west were interrupted. North of this
area the power system was intact including the interconnections to Norway and
Finland. Supplies were not primarily interrupted in the stockholm area. Some
sensitive equipment reacted however to the low voltage level and transients,
leading to a few irregularities in traffic control systems and telecommunications.
Frequency recordings from a 220-kV substation near Stockholm have been
published as part of the officially available documentation of the 2003 blackout
in Sweden. In addition, recordings from three phasor measurement units (PMUs)
temporarily installed on the island of Oland (some 400 kilometers south of
Stockholm and east of mainland Sweden) are also available. Figure 2-12 shows
the location of the three PMUs.

2 - 22
Figure 2-12. The Sites of the Three Temporarily installed PMUs

Figure 2-13 shows the frequency recorded in the 220-kV substation Beckomberga
close to stockholm and in three 50-kV substations on the island Oland. L6ttorp is
situated in the northern part of the island, tlnscnkon in the central part of the
island where 130-kV connection from mainland Sweden is connected and
Degerhamn in the southern part of the island.

2 - 23
Recordings in Sweden 2003-09-23
1- Linsankan - Lottorp - Degerhamn - Beckomberga I

I t--
50.2

50.0 I I I
I~
........
N
~ 49.8
I I I,
I.....
>.
(J
A/ I I I

-\
~ 49.6
~
0-
Q) 49.4
~
u.
49.2

49.0
300 320 340 360 380 400 420
Time [s] after 12:30

Figure 2-13. Frequency Recordings from the 2003 Blackout in Sweden

The double busbar fault caused the disconnection of two big generating units at
about 300 seconds after 12:30. The recorded minimum frequency is about 49.3
Hz during the initial phase of the blackout . The frequency increased at the
voltage collapse during the final phase of the blackout.
Figure 2- 14 shows the frequency at four different locations during the init ial phase
of the blackout . The average value of the lo c a l frequency recorded on the
island O land is very similar to the system frequency recorded close to Stockholm
some 500 kilometers apart . There is, however, a difference in between the system
frequency recorded on the 220-kV level and the local frequency on the 50- and
10-kV level during the first second after the busbar fault. The rate of change of
the system frequency recorded on the 220-kV level is about 0.12 Hz/second while
the rate of change of the local frequency recorded on the 50- and 10-kV level is
about 0.25 Hz/second. The system frequency recorded on the 220-kV le v e l is
smoother than the local frequency because of different frequency transducers
and different type of low pass filtering .

2 - 24
Recordings in Sweden 2003-09-23
1- Linsankan - l.ottorp - Degerhamn - Beckomberga
50.2 .- .----- - - - - - - - - - --..,.-- - - - - - - - - - ---,

50.0 1- - - - - - - - - - - -+-- - - - - - - - - - - 1
N
~ 49 .8 ' - It '-'--:c_ -- - - - - - - - + - - - - - - - - - - - -l
>.
(J
l: 49 .6
OJ
~

g 49 .4
l-
LL
49 .2

49 .0 1 - - - - - - - - - - - - - - ' - - - - - - - - - - - - - - - '
300 305 310 315 320
Time [5] after 12:30

Figure 2-14. Frequency Recordings from the Initial Phase of the Blackout

There are a number of spikes in the frequency recorded on the island Oland and
Figure 2-15 shows two of the recorded frequency signals with better resolution.
The spikes at t=300A seconds and t=301.3 seconds are recorded at two different
location 50 kilometers apart. The switching of circuit breakers during the final
phase of the voltage collapse is associated with an abrupt change in absolute
phase angle. It has been concluded that the step in phase angle causes a spike
in frequency derived from the rate of change of the absolute phase angle of the
busbar voltage. The integral of the frequency pulse is related to the size of the
step in phase angle.

Recordings in Sweden 2003-09-23


- Lin5ankan - Degerhamn
50.2 .----------,-----~--------_:__---__,

50.0
N '
~ 49 .8
>.
(J
l: 49 .6
OJ
~

g 49.4
l-
LL
49.2

49.0 1-- ~ _;__--------....,.._-------'


300 301 302 303 304 305
Time [5] after 12:30

Figure 2-15. Details of the Frequency Recordings from the Initial Phase

2 - 25
Figure 2-16 shows the voltage recordings from the 2003 blackout in Sweden.

Recordings on Oland 2003-09-23


- 50-kV Linsankan -1 OkV Lottorp - 50kV Degerhamn
1.25

~ r-v-
I
I

\ 1
I
1.00 I

'5'
~ 0.75
Ql
Cl I
ctl I
~ 0.50 --
o
>
0.25

0.00
300 320 340 360 380 400 420
Time [s1 after 12:30

Figure 2-16 . Voltage Recording from the 2003 Blackout in Sweden

There are three distinct dips in the recorded voltages. The first is associated with
the double busbar fault on the west coast of Sweden . The second, at t=312
seconds, is associated with a local voltage collapse in the subtransmission system
and the third, at around t=388 seconds is associated with the voltage collapse in
the 400-kV transmission system . Figure 2-17 shows the frequency recordings from
the final phase of the blackout .

Recordings in Sweden 2003-09-23


- l.lnsankan - t.ottorp - Degerhamn - Beckomberga I
50.2

N
50.0
~v
~ 49 .8
>.
(J
49.6
::::l

...go 49.4
u,
49.2

49 .0
I
380 385 390 395 400
Time [s1 after 12:30

Figure 2-17. Frequency Recordings from the Final Phase of the Blackout

2 - 26
The frequency recorded on the 220-kV level close to stockholm and the
frequency recorded on the 50- and 10-kV level on the island Oland increased by
0.4 Hz during the last phase of the blackout. The voltage decay in the southern
part of Sweden caused a load reduction because of the voltage dependence
of load. The load reduction was so big that it affected the power balance in the
synchronous Nordel system and caused the frequency increase.
2.3.3 Summary
In this subsection a brief summary has been given of the load frequency control
strategy in the Nordel system. In addition, some examples have been given that
clearly show the system response for some major system disturbances that have
resulted in large changes in system frequency.

2.4 North Dakota / Minnesota Generator Response and Exposure - The


June 25, 1998 Northern MAPP Disturbance
2.4. 1 Overview
The June 25, 1998 disturbance in the northern. portion of the Mid-Continent Area
Power Pool (MAPP) resulted in two islanding events. The first event, occurring at
2:21 :56, yielded an island containing North Dakota, northern and central
Minnesota, northern South Dakota, and Manitoba and Saskatchewan.
Frequency in the island reached 61 .10 Hz.
At 2:25:47, a second island formed within the initial one, containing just North
Dakota and northwest Minnesota. The frequency inside this smaller island
reached 62.29 Hz, and settled to 61 .3 Hz, where it stayed for about 25 minutes.
The remainder of the large initial island rapidly resynchronized with the Eastern
Interconnection following formation of the North Dakota island. The frequency
was reduced in the second island and it was reconnected to the Eastern
Interconnection at 3:03:39.
This report documents the performance of on-line generators within the North
Dakota island, as well as noting the negative impact of the prolonged exposure
to the high frequency.
2.4.2 Governor Response in the North Dakota Island
2.4.2. 1 Background Information
When load is changed abruptly on a generator, its natural response is to
accelerate if load is removed, or decelerate if load is added. The amount of the
change in speed is limited by the governor of the machine within several
seconds. The governor is designed with a frequency versus power output
characteristic referred to as droop. By definition, the droop is the percent
change in frequency required to cause the governor to move its valves from the
machine's "no-load" to "full-load" condition. As an example, if a single fully
loaded generator with a 5% droop setting on its governor is operating at nominal
frequency and serving an islanded pocket of load, and suddenly half of its load

2 - 27
is lost, it will settle at a frequency 2.5% above nominal. On a 60 Hz system, this
frequency will be 6 1.5 Hz.
Because of the natural generator inertial response, the droop characteristic is
sloped such that loss of load yields a final machine speed greater than normal,
and addition of load results in a lower machine speed than normal. The
generator will reach a steady state speed when its power output is matched to
the real power load applied to it.
In addition to speed regulation, the generator governor is the control
mechanism that is used to set the generator's real output power. As a result,
control actions that are taken to adjust the generator power also alter the
generator's response to frequency deviations. Raising the governor load
reference to increase power output will raise the system frequency unless a
corresponding load reference reduction is made on another generator within
the region of electrical interconnection. Lowering the load reference will have
the effect of lowering system frequency if no corresponding increase is made
elsewhere. Because of this fact, non-coordinated efforts to change generator
power output during an off-normal frequency event will change the system
frequency, either towards or away from the normal level, aiding or complicating
restoration efforts.
Since there is some frequency sensitivity of load, the real power requirement to
serve the load will change in an off-nominal frequency event. The load level
increases by approximately 1% for every 1% increase in frequency. This fact has
a mitigating effect on the generator's initial acceleration or deceleration, aiding
the governor in restoring the generation and load balance, and yielding a
smaller overall deviation in frequency than there would have otherwise been.
The June 25, 1998 disturbance involved loss of loading on the North Dakota
generators, so it resulted in over-frequency in the island rather than under
frequency.
2.4.2.2 NERC guidelines for operation of turbine-governors:
The following points from NERC Policy 1C refer to generator governor
requirements.
1. Governor installation: Generating units with nameplate ratings of 10
MW or greater should be equipped with governors operational for
frequency response unless restricted by regulatory mandates.
2. Governors free to respond: Turbine-governors and HVDe controls,
where applicable, should be allowed to respond to system frequency
deviation, unless there is a temporary operating problem.
3. Governor droop: All turbine generators equipped with governors
should be capable of providing immediate and sustained response to
abnormal frequency excursions. Governors should provide a 5% droop
characteristic. Governors should, as a minimum, be fully responsive to
frequency deviations exceeding 0.036 Hz ( 36 mHz).

2 - 28
4. Governor limits: Turbine control systems that provide adjustable limits
to governor valve movement (valve position limit or equivalent) should
not restrict travel more than necessary to coordinate boiler and
turbine response characteristics.
2.4.2.3 Analysis of Recorded Data:
North Dakota generation and tie flows are recorded by MAPP at regular
intervals. These recorded values were used to calculate the loading of the
islanded generators before and after the island was formed. The goal of the
calculations is to show the expected frequency during the islanding event,
assuming no machines tripped, and all machine governors were in service with a
5% droop characteristic.

F.final = F.initial + Droop x 60 Hz x [(P.initial- P.final) / P.gross]


Assuming the following variable definitions:
F.final operating frequency during disturbance due to governor
action
F.initial operating frequency prior to disturbance
Droop governor droop, assumed to be 5%
P.initial gross real power output of generation before the
disturbance
P.final gross real power output of generation during the
disturbance
P.gross gross rating of on-line generation

F.final = 60 + 5% x 60 x [(4,064 MW - 1,867 MW) / 5,530 MW]


= 60 + 1.192
= 61.19 Hz

For the 5530 MW of gross machine capacity of the 23 generators that were on-
line, the gross machine loading went from 4,064 MW to 1,867 MW when the
island occurred. The 5% droop calculation on the composite of these machines
(see above) indicates that the frequency should have gone from 60.0 Hz to 61.2
Hz in the island.
Since the actual frequency reached 62.3 Hz, it is evident that the response of the
machines and their operators was not as desired. In fact, two large generators
had tripped off prior to the frequency reaching this level, so the frequency
should have peaked substantially below 61 .2 Hz.
Figure 2-18 shows a plot of system frequency for the event. Recordings of real
power output from each of the generators were compiled, and correlated with

2 - 29
system frequency. Example graphs from this effort are shown in Figures 2-19 and
2-20. The graphs contain a solid line indicating the droop characteristic based
on the initial operating level of the generator, and a dashed line for the droop
characteristic at the generator's rated MVA. Ideally, the actual operating points
would fall along the solid line droop characteristic.

62.S .,...-- - - - - - - - - - - - - - - - ---.,

(l ~ . O

....
~ (, 1.0
~

- Initial Isbn d Iormation


60.0
c
c, r l -r,
.;. 'r ,
~
'" 'r ,
H . '1' ':.
c.
~.

c
r-
" " '"
0 ' I r-

Figure 2-18. Recorded System Frequency in North Dakota


during the June 25, 1998 Event

6.1 ,00 . . . - -- - - - ______,_______,_______,_______,_______,_______,_______,_______,_______,_______,_______,_______,_______,_---.,


I ~ :Ic l lla l ;~;;~ ;::;i ng;)"i ll i ~--_...
62,50 .. '.... I' " .
5"" droop. rated machine MVA
5" " d roop , initia l
- 62,00 - 11.7"" d, " np with Iilllll ( frcq . hias co lltro l)
--."
~

~ ' 6 1.50
::
:i: 6 1.00
.~

60 .50

n 100 200 ,10 (J 4()U 50 0 (lOU


Gene rator Gross 0 111/1111 ( , IWi

Figure 2-19. Example Plot of Actual versus Expected Response of a Machine during the
June 25, 1998 Event. Droop is much higher than 5% on this machine.

2 - 30
() .~ .O .r-- - - - - - - - - - - - - - - - - - -----,
. act ual op era ting point
6 ~ .5 . ... sn" droop , rated machine MVA
- 5" " droop, in itial

--
.;-; () ~ .ll .
~

~. (\1 5
~

c: i.o
~

(;0 .0 .'--- - - ,..--- - - ,..--- - ""'4-_ - - ,..--_ ...:.......-_ ----!


1I I II 20 30 so
Gcncn uor SC I 0 11I1' 11I 1.1{/J )

Figure 2-20. Example of Machine Response during the June 25, 1998 event.
This example shows evidence of load reference increase
during a time period following the event.

From observation of the graphs, the following reasons why frequency went as
high as it did are apparent:
1. Governors not operational

2. Governors with greater than 5% droop setting

3. Governor load references raised during the event

Governor droop can be detected visually by observing the droop plots, and
noting the slope of the line followed by the operating points throughout the time
period. Non-operational governors were evidenced by a complete lack of
correlation between frequency and generator power output.
Table 2-3 summarizes the governor performance of the islanded generators.

Table 2-3. Identified Problems with Unit Governor Response


Number of Generators Governor Performance
8 Operated properly with 5% droop
2 Appeared to function properly with 5% droop
6 Appeared to function properly with a droop setting
substantially different that 5% (varied from 1% to 14.7%)
5 No discernable governor response, either due to low
resolution of recorded data, or there was no alignment
of operating points along a possible droop curve
2 Generator was operating as a condenser throughout
the disturbance (real power output of 0)

2 - 31
In some cases, uncertainty is noted in the table because the recorded values did
not clearly reveal what the droop characteristic was. This problem could have
been due to lack of coordination between the governor and boiler controls on
steam turbines, or because the recording mechanism was too slow or lacked
accuracy. Times were adjusted on recordings where it was clear that the
recorded times were in error. Time alignment problems would also contribute to
difficulty in correlating generator output with frequency.
A prevalent problem driving the frequency up appears to have been raising of
the governor load reference. Load reference changes could have been made
by Automatic Generation Control, or manual operator action at a control center
or generating station, or by the plant control system. The most likely cause of the
change is manual operator action. This could occur if the plant or system
operator was unaware of the events taking place on the power system, and was
attempting to keep the power output from the generator at its scheduled value.

2.4.2.4 Possible Influence of Generating Plant control Systems:


It is possible that governor action was hindered on some generators by the
action of coordinated boiler/turbine controls. These controls measure pressure,
temperature, and flows in a boiler and turbine, and make adjustments as
necessary to maintain generator output at the desired real power level while at
the same time maintaining boiler and turbine inlet steam pressure. A frequency
bias feature in the controls allows the generator real power set point to be offset
by a factor reflecting frequency deviation from normal. The offsetting factor
should be programmed to match the governor droop. The feature keeps the
coordinated controls from counteracting the influence of the generator
governor.
This matter is discussed in the April 1988 issue of the PTI Newsletter, in an article by
D. N. Ewart entitled "Who's Watching the Frequency These Days?" According to
Mr. Ewart, the frequency bias feature is not well understood. The following quote
from the article is pertinent.
"Unfortunately, the role of the frequency bias circuit is very much
misunderstood. Some take the view that the turbine-governor's role is to
protect the turbine, and the boiler control's role is to control MW, and
therefore frequency bias has no function. In some instances frequency
bias options have not been purchased, and in others the circuits have
been discarded, or their place has been taken by "home-brew" circuits to
accomplish other special functions. A sampling by PTI indicates a fairly
pervasive tendency toward non-use of frequency bias."
"Is this causing a problem? We think it might be. We hear expressions of
concern that under normal conditions, frequency response to routine
system disturbances is larger than it should be, perhaps as much as a
factor of two."

2 - 32
2.4.2.5 Over-Frequency Exposure
No major equipment damage was reported as a result of the disturbance. The
prolonged exposure of the generators in the North Dakota island to high
frequency may be cause for concern, however. The longer blades of a typical
low-pressure steam turbine have natural resonant modes that fall within the
range of the lower harmonics of the normal operating speed (3600 RPM).
Constructing the blades to withstand continuous operation at a resonant mode
is not practical, so they are specifically designed so that they will not be at or
near resonance at normal operating speed. As the operating speed of the
turbine varies from normal, the blades are exposed to impulse frequencies that
approach their natural resonant modes, and excessive vibration results.
Prolonged operation at such near-resonant frequencies will likely cause
premature fatigue-induced failure of the turbine blades. A detailed explanation
of this issue is found in [18].
The frequency deviations experienced during the disturbance are listed in Table
2-4 below along with the manufacturer's recommended limits from one of the
turbines that was in the island.

Table 2-4. Comparison of Over-Frequency Exposure With One


Particular Turbine Manufacturer's Specified Limits - June 25, 1968
Frequency Deviation Related Recommended Maximum Actual Over-frequency
from 60 Hz Range of Life-time Exposure Exposure on June 25, 1998
Frequencies in the ND Island
+/- 0.5 Hz 59.5 - 60.5 Hz No limit N/A
+/- 0.5 - 1.5 Hz 58.5 - 59.5 Hz 50 minutes 37.8 minutes
60.5 - 61 .5 Hz
+/- 1.5 - 2.1 Hz 57.9 - 58.5 Hz 10 minutes 1.2 minutes
6 1.5 - 62. 1 Hz
+/- 2.1 - 2.6 Hz (approx) 57.4 - 57.9 Hz 2 minutes (approx) 1 second
62.1 - 62.6 Hz

A protective device at one generator, designated as an Overspeed Protective


Controller (OPC), operated numerous times beginning at 2:25:44. The control
action taken when this device operates is to close the steam turbine intercept
valves, which removes the input power from the turbine. This introduces a large
short-term change in power output from the machine. The setting on the OPC is
103% (61 .8 Hz). Its operation log shows action coinciding with the peaks of the
large frequency oscillations that occurred at this time. The oscillations are shown
graphically in Figure 2-21.

2 - 33
There are a number of generators in North Dakota that have the same model of
control system as the one described above. Since the setpoint on this protective
device is likely established at the factory, these other generators may have been
acting in the same way. The magnitude of the frequency oscillations suggests
that they were.

(,25 ',--- - - - - - - - - - - - - - - - ---,

.
:; G!5

(,0,5

-r ,
r,
ee:
sr ,
'n
-:t:
tr .
S'
v:
-
-r,
-:i
~,
Ir :
~
.~ ,-,
"

r 'l
r:'! <'I (""I
~! ~t <"', <"','
<""~I <"', <",., <""l (,;1 <"", r ,. <""l

lime (hour:mil /Wei

Figure 2-21. Large Frequency Operations during the June 25, 1998 Event.
Generator overspeed protective controller operations were occurring during this time.

2.4.3 Follow-up Recommendations


The following is a list of recommendations that came out of the postmortem
analysis of this event.
1. Generator governors should be tested routinely. They should be set to
match the guidelines contained in NERC Policy 1C. The older
mechanical governors can be tested by observing and recording
governor valve deviation at various speed points as the
turbine/generator shaft is slowing down after being taken off-line.
At least one of the newer software implementations of governors has
no provision for testing the governor operation. Some means of testing
needs to be developed. A possible recommendation is that a test
device be developed for such systems so that routine test ing can be
performed. One way this can be done is by artificially varying the
machine speed ind ic a tio n to the control system with an auxiliary circuit
and observing governor valve response or generator output. This
would be most desirable to do while the machine is carrying load,
perhaps close to the minimum level to reduce the risk of machine
damage if problems develop with the test system. The injection of
such artificial test signals can be programmed into the turbine controls,
if the testing staff have an intimate knowledge of the control system .
However, the recommendation here is to make such a feature readily
available to the operator.

2 - 34
2. Control center and generation plant operators that control generator
output via AGe or local controls should receive training on how
governors work. For the particular governor implementation on the
generators they control, they should know what the high/low limits of
governor response are, and in which control modes the governor is
disabled. They should also be instructed not to attempt to disable the
governor or change the generator power output during off-normal
frequency conditions until the problem is diagnosed, and corrective
action is prescribed by the relevant control center personnel.
3. Generating plant operators should have a frequency meter in clear
view, and possibly an alarm level set to give positive indication that
the generator is operating in an abnormal frequency condition.
4. steam turbine generation owners should review the impact on their
units due to any prolonged over/under-frequency exposure following
a major grid disturbance. As such, they should consider any necessary
actions to protect their units from significant loss of life in the blade
structures due to such exposure - e.g. revising protective relay settings.
5. Turbine/boiler installations using coordinated controls should verify that
the frequency bias feature is in service, with a setting that yields the
same droop response as the governor.
6. A lack of accurate, digital format recordings was a serious limitation in
the ability to determine how many of the generators responded.
Digital disturbance recorders with GPS time clocks should be
considered for installation at generating plants to address this
shortcoming. To facilitate future analysis, they should record generator
real power output and system frequency.

2.5 Regression Analysis of The Eastern Interconnection


This brief section presents regression analysis that looks at many system events
(on the US Eastern Interconnection) for events that resulted in a loss of
generation of 700 MW, or more.
Figure 2-22 shows a histogram plot of the distribution of events versus the
corresponding MW/O.l HZ13 response of the interconnected system.

13 It is recognized that the units 'MW/O.IHz' are no longer used. The more commonly used units in present
practice are MW/Hz or MW/mHz. However, the plots presented here were only available in these units.

2 - 35
Histogram of Eastern Interconnection Frequency Response ('94- June '04)
40
r--
r--

30 -

....IIIe -
Cll
r--
~ 20
'0 .--
'*t
-
I--
10
-
-Hl
r--

o r---l I I I I r---l
1800 2400 3000 3600 4200 4800 5400
Frequency Response (MW/O.1Hz)

Figure 2-22. Plot of Number of Events versus the Response Characteristics


of the Event for Events between 1994 and 2004
Figure 2-23 shows the same events, this time plotting the d istribution of resultant
frequency excursions. Finally, Figure 2-24 shows the plot of the system MW /0.1 Hz
governing response versus the event year. The regression tested for a decline in
frequency response since the study by Jim Ingleson and Makarand Nagle (since
March 1999) [36]. The regression also included a dummy variable that checked
for differences in the summer months (June-August). There is evidence to
support the assertions that frequency response has declined (-145 M W/0.1 Hz)
since the 1999 study and that there is more response in the summer months (491
MW /0.1 HZ). Figure 2-24 shows a clear indication of the decreasing trend in total
system governing response in the last several years .

2 - 36
Histogram of Frequency Excursions
70 -
r--
60

50 f--

...
III

~ 40
-
~
....
o 30
'**'
r--
20 -
r--
10

o ~ IL ~

-0.08 -0.06 -0.04 -0.02 0.00 0.02 0.04


Delta Frequency (pre to post disturbance)

Figure 2-23 . Plot of the Distribution of Events with regards


to the Resultant Frequency Excursion

Boxplot of Frequency Response vs Year

5000

CII
III
e 4000
0
Q.
III
CII
0::
>
c: 3000
u
CII
::::J
cr
CII
a:
2000

1000 '----,----,---,-------,-----,----,-----,---r----,-----y--..,---'
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
Year

Figure 2-24. Plot of MW /0.1 Hz Governing Response of the System versus Event Year

2 - 37
2.6 Governor Goals and Characteristics in UCTE
2.6.1 UCTE Requirements
UCTE14 defines a common set of requirements for generating units to be
connected to the UCTE transmission svstem". It sets up a common framework for
grid connection agreements between transmission system operators (TSO's) and
power station operators. TSO's are entitled to impose additional or more precise
requirements on the generating units when needed for system operation.
The requirements defined by UCTE shall apply to all thermal (e.g. coal, gas-fired
and nuclear) and hydro generating units above a minimum size of capacity. This
capacity threshold can be defined by each TSO individually (generally the
threshold is 10 MVA).
To other types of generating units (e.g. wind generators), UCTE requirements for
conventional generation shall apply as far as possible, especially if their purpose
is to avoid risks for system security (e.g. voltage ranges, reactive power supply,
behaviour in case of perturbations to the system like short circuits, etc.).
Additional requirements considering the specific characteristics of wind
generators can be defined by each TSO and are currently under development
by several TSO's. UCTE requirements should also apply to generating units
connected to distribution systems as well.
UCTE recognizes that generating units have the capability to significantly
contribute to system security by providing so called ancillary service to the TSO,
who provides the system services. Furthermore these ancillary services have to
be provided on a mandatory basis to a certain extent meeting the relevant
minimum technical requirements defined by each European TSO. Moreover
appropriate dynamic behaviour of generating units, protection levels and
control facilities are necessary in normal operating conditions and in a range of
disturbed operating conditions in case of perturbations to the system or during
system restoration in order to preserve or to re-establish system security and
equipment integrity.
It is also prescribed that the technical requirements to generating units shall be
accountable and measurable to allow the TSO's to check and verify their
fulfilment according to established good practise during commissioning of new
generating units as well as repeatedly during operation or after maintenance.

14Union for the Coordination of Transmission of Electricity.


15UCTE Synchronous Area - is a part of a synchronous area covered by interconnected
systems/TSOs which are members of the association. Different UCTE Synchronous Areas
may exist in parallel on a temporal or permanent basis.

2 - 38
The relevance to UCTE system security is characterised by criteria that have
system-wide impacts and are intended to prevent system-wide emergency
situations or even blackouts. Especially frequency criteria such as over- or under-
frequency disconnection of generating units, or frequency limits for indefinite
nominal active power supply can be identified in this category. Another issue to
be mentioned is wide-scale inter-area oscillations that can be mitigated
appropriately by power system stabilizers or comparable installations.
Moreover, each European TSO needs to set prescriptions which are relevant for a
controlled system operation. This category is characterised by requirements that
are needed to ensure system security on a national basis. The requirements
support to achieve a comparable security level within the UCTE interconnected
transmission system. Typical requirements for this category are reactive power
supply conditions, operation in emergency situations such as the capability to
withstand grid faults without tripping and also the behaviour of generating units
after disconnection from the grid to enable a fast reconnection in the restoration
phase.
Finally, other rules should regard the security of the local/regional grid. These
requirements cover the normal operating conditions of the regional grid to which
the generating units are connected. Examples are settings of control and
protection equipment to coordinate grid and unit protection and other
equipment design parameters. Some of these aspects can need coordination
between neighbouring TSO's.
Among the requirements that are relevant for the European system security,
UCTE considers that unit tripping, or power reduction, are forbidden during
power swings and frequency oscillations commonly recorded in the frequency
range of 0.2 Hz to 1.5 Hz.
Similarly, automatic disconnection of power units is forbidden in the frequency
range from 47.5 Hz to 51.5 Hz. This applies for generating units directly
connected to each TSO grid as well as for generating units connected to TSO
customer grids (e. g. industrial sites). If specific disconnection concepts are
considered necessary, e.g. to secure industrial processes, they shall be agreed to
by the TSO in advance.
Prescriptions more relevant for the national security level include transient
stability requirements for short circuits near the generation unit, with voltage dips
below 85% of the nominal voltage. In these case, the disconnection of power
units is forbidden, if faults are cleared as fast as possible according to each TSO's
standard fast fault clearance times taking into consideration the short-circuit
power level at the grid connection point.
Policy 1 of the UCTE Operation handbook describes load-frequency control and
performance.
The minimum continuous load change rate between Pmin and Pmax is 1%
Pmax/min. Each TSO is entitled to require higher load change rates taking into
consideration the type of generating unit.

2 - 39
The capability to provide primary control is mandatory for generating units
above a minimum capacity to be defined by each TSO. Each TSO is entitled to
define a minimum primary control range for generating units in terms of Pmax.
The primary control range must be adjusted as instructed by the TSO when
participating in primary control.
Among other subjects, it is clearly mentioned that a common goal is that
frequency deviation should not exceed 20 mHz for the European
interconnected grid, (called first synchronous zone). That is to avoid calling up of
primary control in undisturbed operation. Conversely, primary control is
activated if the frequency deviation exceeds 20 mHz. Additionally, the set of
requirements for primary control is based on the aim that the maximum quasi
steady-state frequency deviation must not exceed 180 mHz and the maximum
permissive dynamic (instantaneous) frequency deviation must not exceed 800
mHz.
These general requirements are related to the maximum power/load unbalance
that should be handled by primary frequency control, starting from an
undisturbed operation of the UCTE power system. The maximum single incident is
set at 3000 MW of power-load unbalance, depending on the size of the
synchronous oreot- and the size of the largest generating unit or generation
capacity connected to a single busbar.
The total deployment of the primary reserve should start a few seconds after the
frequency disturbance and should end 30 seconds thereafter, with the
additional constraint that 50% (or less) should be provided by the first 15 seconds.
The total contractually agreed primary control power shall be fully activated at a
quasi-steady frequency deviation of 200 mHz within 30 sec and supplied for at
least the next 15 min. The same rate of power change shall apply in case of
minor frequency deviations.
In addition, the accuracy of frequency measurements and the insensitivity of
controllers should not exceed 10 mHz. Nevertheless, the goal for new units is
that the accuracy of frequency measurement shall be below 10 mHz and the
measurement cycle must be in the range of 0.1 sec to 1 sec.
For primary control, the insensitivity range shall be as low as possible and should
not exceed 10 mHz. A flexible deadband and its settings can be agreed
between the TSO and the power station operator.
The total size of primary reserve is the same size of the maximum incident. UCTE
rules assign each country/regulating block a percentage of the total amount of
primary reserve, as a function of the country/block dernond." However, it is
requested that the total primary control must be fully activated in response to a

16 In this case, the 3,000 MW value is for the power systems of the European countries

except United Kingdom and Russia. The system load typically ranges from 150 GW off-
peak to 300 GW peak.
17 For example, the given percentage is 11.8% for Italy.

2 - 40
quasi steady-state frequency deviation of 200 mHz.18 According to the above
requirements, this translates to a droop characteristic of the UCTE synchronous
zone of 15,000 MW/Hz.
2.6.2 The Italian T50 Requirements and Experience
2.6.2. 1 Governor Requirements
The Italian TSO requires that every unit over 10 MVA of rated power contributes
to primary frequency regulation according to the following rules, except for
those who are naturally without regulating capacity, such as, for example
renewable-energy based power plants.
2.6.2.2 Characteristics of Speed Regulators
Each unit must be equipped with a governor of which the load reference signal
may be varied from 0 to 100% of the nominal load within a maximum time of 50
s. The regulator must be able to correctly operate even in disturbed auxiliary
supply voltage conditions (for example, during transmission network transients).
For each unit performing primary frequency regulation, the governor must have
the following functional requirements described below.

Droop Settable between 2% and 8%


Maximum speed tolerance 0.02%
Maximum insensitivity zone +/- 10 mHz

The governor must guarantee indefinite stable group operations, for any
frequency between 47.5 Hz and 51.5 Hz, and with any load between the auxiliary
service load and the nominal power of the unit. Furthermore, correct operations
down to 46 Hz must be guaranteed for limited amounts of time (several
seconds).
Thermal units must be able to regulate frequency even when operating on an
isolated grid portion so as to return and keep frequency at nominal value 0.25%,
just to allow grid reconnection.
The droop value is set according to the unit characteristics, the plant connection
to the transmission grid and the possible use of the plant in the grid restoration.
Recently, the Italian TSO, for units above 10 MVA of rated power, defined that
the droop must be set to:
5%, if the unit is a conventional thermal unit. This also applied to
combined-cycle power plants (including both steam and gas turbine
response).
4%, if the unit is a hydro unit.
Additionally, the Italian TSO prescribed that the deadband must be set at a
maximum to:

18 This is a requirement that does not consider the self-regulating effect of the load

assessed to be 1% for a 1% frequency drop.

2 - 41
10 mHz, for each kind of unit except gas turbines.
20 mHz for gas turbines.
For old units, independently of characteristics, the combination between the
intentional deadband and the regulator insensitivity should not exceed 30 mHz.
2.6.2.3 Band of Primary Reserve
The TSO requests that each production unit must provide at least 1.5% of its
efficient power'? as primary reserve during a system event, if connected to the
continental UCTE grid. Otherwise, for the power systems that are always
(Sardinia grid) or temporarily (Sicily grid) disconnected from the UCTE grid, the
primary reserve band must be at least 10% of the efficient power of each unit.
2.6.3 Power Plant Behavior During Major Disturbances:
A Brief Description of the 2003, 28th September Black Out
On September 28, 2003 the Italian power system was operating under N-1
security conditions. At 03:25:42, a separation between UCTE's system and the
Italian system occurred. The Italian power system remained separated from the
European system with a power deficit of about 6.7 GW in comparison to a total
internal demand of 27.2 GW.
The negative imbalance between load and generation caused a frequency
drop, followed by the operation of primary control of the generating units, the
disconnection of the in-service pumping units and the load shedding in high-
and medium-voltage substations.
Before reaching the state of blackout, the Italian power system experienced a
frequency transient for about 2.5 minutes, starting from its separation from the
European system. The trend of such a transient is affected by the response of
grid protection systems, defence systems, regulating systems and by the tripping
of various generating units. The primary control of in-service generating units
reached its maximum value, about 1.465 GW a few seconds after the event.
Nevertheless in this time interval, frequency fell 1.5 Hz.
The constant decline of the system frequency was due to the loss of major units
(larger than 50 MW) totaling 4,132 MW and of about 3,400 MW of power input on
the 150-132 kV and distribution grids, totaling 7,532 MW. A small number of
thermal units successfully completed their load rejection procedures and thus
remained on their house load. This was useful for the subsequent restoration
phase.
At the end of the 2.5 minutes (Figure 2-25) following the separation from UCTE's
grid (Figure 2-26) the unbalance was about 900 MW. This was due to loss of
power input on the grid amounting to about 14,206 MW, whereas the
disconnected load was equal to about 10,900 MW and the combined
contribution to primary regulation from load and governors was about 1,400

19 The Italian TSO defines the efficient power of a unit as the real power that a thermal
unit can continuously supply or that a hydro unit can supply for a specific period
according to the capacity of the reservoir.

2 - 42
MW 20 Therefore, owing to the imbalance between generation and loads,
frequency fell below 47.5 Hz, inducing the tripping of all generating units still in
service and the consequent loss of the load, except for one supplied by a
thermal power plant in southern Italy.

FREQUENCY
IRome)
50 ,500

50,000

49,500

49,000

48,500

= 48,000

47,500

47,000

46,500

46,000

45,500
o I'- "<T .... 00 I.() N ~ ~ M 0 h ~ _ w ~ N ~ ~ M Ol'- ~ .... 00 ~ N m ~ M Ol"-
OO .... NN M ~ ~ ~ 0 ~ ~ N M M ~ ~ ~ 0 _ N N M ~ ~ ~ 0 0 _ N M M
~ ~ ~ ~ ;t ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ N N N

~ ~ ~ ~ ~ ~ ~ ~ 8 8 8 ~ 8 g g 8 8 8 2 8 ci 8 8 8 8 8
...., ..., ..... M M M M M M
o 0 0 0 0 0 a a a
or.
(~ . - 2,5 minu tes (~

Figure 2-25: Frequency Transient Experience by the Italian Power System,


after the Disconnection from the Rest of the European UCTE Grid

20The purpose of the Italian automatic under-frequency load shedding plan is to arrest
the frequency above the value of 47.7 Hz, to avoid the intervention of the minimum
frequency protection of the generators that intervene at 47.5 Hz. The load shedding is
performed by means of relays that firstly shed pumps, if they are running , from the setting
of 49.7 Hz to 49.1 Hz. In the range from 48.8 Hz up to 47.7 Hz there are typically about 12
steps of 0.1 Hz each (the interval and the number of the steps change within the Italian
areas depending on the typical shape of the load and the network characteristics).
Each step sheds decreasing percentage (from 5% down to 3 %) of the load of the
related area. The first steps of the load shedding plan are set to act both at a minimum
value of frequency and at a certain derivative value of the frequency. The total load
under automatic load shedding is about 50% - 60 % of the national demand. From the
theoretical point of view tha t load was about 10,000-12,000 M W at the moment of the
Italian blackout. Additionally there was 3,280 MW of pumping storage. It should be
considered that at the time of the blackout not all the (daily) industrial loads where
available. Moreover a part of the load was not under the load shedding defense system
because of the characteristic of the residential load during the night mainly: hospitals,
railways, important cities, etc . At the end the percentage of the relays that properly
worked was about 90% and the domestic/industrial load shed was about 7,680 MW.
Then, including pumps, the total load that was shed by the intervention of the automatic
load shedding was about 10,960 MW .

2 - 43
f, U, Q Uchtelfangen (0) - Vigy (F) I Heviz (HU)
28.09.2003; 03:25 (UTC)
6(, 350 ... ....................................................................................................................,........ 410
"
~ / "\ f\i.....:r..J- . ~-../'V-v-...r..r~............-..."-~,..r..."'VV'.....-..r...-...I\. ..... U[hV] Q [ U Y3r]

. . .--"'-
1(l1 ~
:m o
5(, 300 ------;-- - --'i~ ~Ir'i V
II
",,0

volta'~ e (llU) ~ ------", r\ fr~uency (DI lise


:
I, 3W
5( ,250 \J! fI f\ f)j,"J I
:1'I II~\"Y 1 ~ 1 --r..\/'~:!)r~\:rlf<4f':f ':00
l::):~~ iJ: :
oW
5( .200
\JubHHl
l')->::~dlf(ilr~t'?,nl :

,j"""....,,'di'., '; (
'I
::1' I \I~I iI II I'I
V
-
frequency (II U)
""'~'VillUW ' ,,
V1.}g-\I "rJll, . 'M ~
I I .' .
2W
izsn

iII
mkH~nl
(,111.1'1 2(0 1[1]0
5(, 150
iI
u .
.1

5(,100
IV lW 1'"
~ re."ctive j"
: I'O" '"r (tllJ ) ~ lCO
5( ,050 ; !
~ _ _); 5(

0 ( ,00 0 t--- ---...::"'\I')"'?--V


\,. ' \'r '.-
....,-rr-."..<.:../ I'
u'vV\;./\,vVv..fVVVVVIJV,f JWVVWNVVVV 0

2nd Jrd I"" lr"pilg , .5)


oli,950
15 20 25 3C 35 40
03 :25:...

Figure 2-26: Comparison among some Electric Variables Recorded


in the UCTE System during the Italian Black Out.
Recordings of frequency (f), voltage (U) and reactive power (Q) for locations
in Germany (D), France (F) and Hungary (HU), on September 28 th, 2003.

2.6.3.1 Causes of Generation Tripping During the Italian Blackout


Here, the main causes of generation loss on the Italian grid during the blackout
are listed and briefly described :
1. High temperature of exhaust gases . Low frequency values force
speed governors to increase gas turbine power outputs in order to
compensate for the frequency drop. For gas turbines, this correct
control can cause exhaust gas temperature to rise above its limit
transiently and if not properly controlled may lead to the unit tripping.
Thus, some gas turbines tripped due to improper temperature control
in some combined cycle power plants.
2. High temperature of boiler heater/re-heater. The increase of gas
turbine power output caused high temperature in the steam boilers in
some re-powered combined-cycle plants. This was a typical
undesirable response of some re-powered steam power plants.

3. low pressure of steam boiler. The long frequency transients drove


governors to completely use the primary reserve of conventional
steam power plants. At the same time, governors were not able to
keep the thermal cycle in stable condition but exploited all the
available thermal reserves.

4. loss of excitation. During the frequency transient the grid local


voltages varied SUddenly reaching very low values. This status caused
the tripping of some excitation systems. Additionally, many generators

2 - 44
supply their own auxiliary services via a step-down transformer. If the
generator output voltage decreases the auxiliary service supply
voltage may no longer be sufficient, and the operation of the whole
generator may be jeopardized.

5. Loss of synchronism. Power swings across the system and frequency


oscillations occurred during the frequency transient. As a result, some
generators lost synchronism with the main grid.

6. Under-impedance relay operation. Large disturbances can cause, on


the transmission lines still in operation, a current increase and a voltage
decrease. This can be interpreted by some generator protections as
an external fault. As a result, the under-impedance protection can trip
to isolate the related generators from the fault.

2.6.4 Italian 150 Requirements for Abnormal Operating Conditions


After postmortem analysis of the blackout, it was decided that points 4 to 6
(above) were due to random events that cannot be completely avoided during
a large disturbance. However, an extensive review of the plant protection
settings was executed after the Italian blackout.
To solve problems listed in points 1 to 3, the Italian TSO has checked with
Manufactures and Independent Power Producers (IPP) the general power plant
performances to ensure that:
New combined cycle gas turbines installed in Italy must support low
frequency transients for an unlimited time without tripping due to
overheating of the boilers and/or high temperature of exhausted
gases (within given limits, i.e. if the frequency remains at or above 47.5
Hz).
steam boilers and re-powered units should limit the initial step increase
of real power due to primary regulation, during low frequency
transients, to a limited value. Specifically, +5% of the efficient power
was a reasonable value chosen by the TSO to limit the sudden power
increase in the first instants after a large frequency deviotion." This
requirement was put into place to prevent these units tripping during a
low frequency event due to high temperature transients in the
boiler/re-heater following a severe event.
2.6.5 Italian 150 Requirements for Normal Operating Conditions
Previously, it was stated that for each unit performing primary frequency
regulation connected to the Italian power system, the governor must have a
droop value between 2% and 6%. Specifically, a value of 5% for thermal units is
desired, independent of the energy source.

21 After that first power increase, given by primary regulation, the local frequency

secondary regulation must govern the unit power supply to solve the frequency transient
with a power rate compatible to the thermal unit characteristics, up to 8% per minute for
gas turbines in combined cycle power plants.

2 - 45
Additionally, the primary reserve should be a minimum of 1.5% of the efficient
unit power. In this case, with a 5% droop, all the primary reserve will be used for a
frequency deviation greater than 57.5 mHz22.
2.6.5. 1 The Combined Cycle Gas Turbine situation
The standard combined cycle power plant is a combination of at least two units,
a gas turbine and a steam turbine (see [1] for more details). The total rated
power of the gas turbine section is generally double that of the steam turbine.
Moreover, during normal plant operation the steam turbine is working at the
maximum power or, conversely, it does not regulate power, having the valve
completely open. This specific mode of operation is called sliding pressure. This
means that typically the gas turbine-governor must provide primary frequency
control, i.e. the initial response to system frequency deviations within the first 30
seconds as requested by the UCTE requirements.
The above means that the gas turbine-governor should have a droop of 3.3%,
which is very low for some governors, acting on an available reserve of 2.25% of
the rated gas turbine unit efficient power, which is considered not acceptable
due to the reduction in plant efficiency.
It was observed that IPPs, in order to get the max profit from power generation,
operate their units at base load (i.e.: maximum exhaust gas temperature or
maximum gas turbine power). Then in order to provide the + 1.5% primary
reserve, they experimented with two different techniques, named peak firing
and over firing:
1. Peak firing. This functionality is invoked manually by the operator; it is
thus independent of the grid frequency deviation. Being manual, it is
not considered a primary regulation. This mode of operation allows
higher temperature for exhaust gases just to allow the unit to produce
more power. However, a timer records the periods of this operation
because it causes a reduction of the hot gas path part life.

2. Over Firing. A specific controller allows the turbine to be fired to a


temperature above its normal base load exhaust temperature by a
limited amount. The limit can be set to allow the turbine power output
to exceed its base load up to a few percent just to provide the
requested primary regulation.

The Italian TSO discourages both of the above mentioned techniques, because
there is no guarantee of either the amplitude of frequency deviations nor its
duration in time.
2.6.5.2 Periodic unit Performance Verification
Recently, the Italian TSO set up a general procedure to perform a periodic test of
governor characteristics and unit performances during normal and emergency

2257.5=37.5+20 mHz, being: 20 mHz the sum of the governor insensitivity (10 mHz) plus the
accuracy of the frequency measurement (10 mHz). UCTE states that frequency
deviations of the interconnected European power system, during the last 3 months of
2004, were below 25 mHz, for 95.4% of the time, with a maximum deviation of 57.1 mHz.

2 - 46
operations. Specifically, every three years each power unit above 100 MVA of
rated power must stand six tests aimed to verify: droop values, ramp rates,
maximum continuous power and the capability to support a -1 ,0 Hz frequency
transient (with a gradient of -0.5 Hs/s) by raising their power up to their maximum
output without tripping (especially due to thermal reosonsl>.
2.6.6 Summary
Here a brief description has been provided on the treatment of primary
frequency control and governing in the UCTE system with some specific
experience and examples from the Italian system. One of the key observations
in the Italian system was related to new combined-cycle power plants and re-
powered steam plants than used the exhaust of gas turbines for additional heat
supply to the steam boiler. The Italian TSO worked with turbine manufacturers to
try to ensure that gas turbines installed in new combined-cycle power plants do
not trip due to excessively fast action of the governors that may drive the
exhaust temperature of the unit beyond its trip point following a major low
frequency event. Similarly, that the sudden increase in heat injection (again
presumably from gas turbine exhaust in re-powered station) into steam turbine
boilers be limited following a major low frequency event to avoid tripping large
steam units due to high temperature transients in the boiler.

2.7 Development of the New Therrnol>' Governor Model in the WECC


2.7.1 Introduction
In the past accurately simulating the frequency response in the Western
Interconnection when large generators and plants trip has been unsuccessful.
Comparisons of disturbance monitoring recordings with the computer simulations
have indicated a wide discrepancy in both the "initial transient dips" and in the
"settling" frequencies. Assessment of the first transient dip is important for load
shedding while the settling frequency is a measure of the responsiveness of
turbine-governors and AGC in the system.
Accurate governor modeling was always high on the WECC modeling list from
the simulations of the 1996 blackouts in July and August in the WECC. In early
2001, in attempting to meet proposed new criteria by NERC for Frequency
Responsive Reserves (FRR), the Western Electricity Coordinating Council (WECC)
determined that the accurate simulation of turbine-governor responses to system
frequency deviations during generator trips was central and critical to
implementing any new requirements.
To further the governor modeling investigation, two separate generation trip
tests, one in the Southwest and the other in the Northwest, were performed on

23 A signal is injected in the governor to simulate that frequency transient which is


coherent to any expected grid accidental separation for the Italian power system (see
the figure of the Italian black out frequency transient).
24 'Thermal' plants include conventional fired steam, nuclear steam, simple cycle gas
turbine, and combined cycle gas turbine plants.

2 - 47
May 18, 2001 in the WECC to determine the response of governors throughout
the system. During the generation trip tests, all Automatic Generation Controls
(AGCs) were switched off throughout the WECC . AGCs operate in real-time to
adjust generation to meet load. Its multiple functions include load-frequency
control, hence when AGCs were switched off, the system frequency response
observed was of governors and load only.
In the first test, 750 MW was tripped in the Hoover power plant in the Southwest.
After the disturbance monitoring recordings were taken, the AGCs were made
operational again to stabilize the frequency and about 20 minutes later, 1,250
MW was tripped in three power plants in the Northwest. Figure 2-27 shows the
frequency response of the second test in the Northwest. Figure 2-28 is a
composite plot of both tests and shows a comparison of simulation results using
existing (incorrect) governor models versus disturbance recordings of the tests.

Summary Plot For FRRte-stA - NorthWest GenT rip @ 1250 MW

os.. .21.0 1_ 13.22:30


f30 .C5 r - - -- - - - - - - - - - - - -- - - - - ------;

II
~. , '1'1.~. ,.jI"''';J\". ."l"'lJ
"J " 't" l d .' . 0." ""~'"
1 ',

60

t - - - - j - - - - + - - -I - , l I . - - j - - - -j - - - --J--- - -

59.90 r-f- - - f- - - f- - - /- - -+- - - ---

598::, -r
(~(" I ~ l l

ss.eo
n l OD "2OJ 30 D 40 0 9 JO 7[0 mo 9 00

Figure 2-27. WECC NW Trip Test on May 18, 2001 of Governing Response only with
AGC Switched Off. Note the pick up of system frequency by AGC 'after' the test

2 - 48
May 18 2001 Test SW and NW Trips - Malin 500 kV Bus Frequency

Base case simulations (existing models)


60 ;" SWTrip NWTrip
, &l /
_..
~
\\\ J- ... ----- ... ~--- - ---------- .
N
I69 .95
\
) . , ... _; .... ..' -----"----------,
~
\

'. ~ay1 !,SW ,150,MW+<T:ri J? .!~ ~o r.~LfJ"


>;
u ~ },*.J""'''' -,,-- ~. I
C
d) ".llAI>_"'"
:::::I
g- 59.9
'-
u,

- May 18 NW 1250 MW Tr ip recording


59.85

~\Jote : AGe was switched off in both generation trip tests


59.8 '----_ _---'---_ _---'- ..l.--..-.._ _- - - ' -_ _-----'_ _- - - - '

o 5 10 15 20 25 30
Time in secs
Figure 2-28. Figure showing the Discrepancy between Existing Model Simulations (blue
plots) and System Frequency recordings (green plots) of the SW and NW System Test Trips
in WECC on May 18. 2001. All AGCs were switched off during the tests.

A simple calculation [6]. using the formula below. indicates that only about 40%
of the governors effectively respond in the real system in the settling time of
about 60 to 100 seconds. If all the governors were responsive in the May 18th
1,250 MW trip test, out of a WECC generation base capacity of 91,000 MW on-
line during the test, the calculated generation pickup for governors with a 5%
droop, for a 0.1 Hz frequency deviation, would be 3,185 M W. Since the actual
pickup was only 1,250 MW , the percentage of 're sp o nsive' governors would be
(1 .250/3.185) or 39%.
R = Droop (regulation) =0.05 pu
D = Damping =0
Setting Frequency Deviation Hz = 0.105
~O) = Frequency Deviation, pu = 0.105/60 = 0.00175 pu
~P = Generation_Pickup, 91,000 MW/base = 3,185 MW

2 - 49
Calculated from:

1
1
-+D
r
Note that this calculation is a simplistic first approach. Load damping, and the
effect of redistributed losses due to different flow patterns in the system after the
trips have been neglected in this simple calculation.
The two tests hence indicated that only 40% of the expected governor response
in the system actually occurred as a result of the initiating generation trip.
However, existing (as in 2001) modeling practice assumes that 100% of governors
respond in accordance with the 5% speed droop governor characteristic. This
results in a significant difference between simulations and actual recorded
system responses.
The principal reason for this large discrepancy is that base loaded and load
limited generators, and units operating with load controllers, are not properly
modeled. These are primarily 'thermal' units, a classification that includes
conventional fired steam, nuclear steam, simple cycle gas turbine, and
combined cycle gas turbine plants. Analysis of the recordings showed that the
hydro units were largely responsive. Investigations indicated that other effects
such as non-linear gate movement, dead band etc have some impact on
simulation results, but a relatively minor one in comparison. In the modeling of
governors, the 'base-load' and 'load controller' operation of units is clearly the
dominant effect.
The new turbine-governor modeling approach correctly represents thermal units
that have demonstrated unresponsive characteristics as "base loaded" units, or
as units with load-controllers (also known as MW power-controllers). Figure 2-29
from Reference [6] compares frequency simulations of the new (correct) model
with the existing (incorrect) model. The comparison is with frequency recordings
during system tests on May 18, 2001 when 1,250 M W was tripped in the Northwest
with AGC switched off system-wide.
The rest of this section describes the development, validation and verification of
this 'new' thermal governor model. The work included the creation of a WECC-
wide system database based on disturbance monitoring and SCADA recordings
of staged tests.
The new modeling approach has been extensively validated against recordings
from three WECC system tests and several large disturbances. The model is now
in use for all operation and planning studies in the WECC.

2 - 50
May 18 2001 Test NWTrip 1250 MW - Malin 500 kV Bus Frequency

60

I Base case (existing models)

:c 59.95
>
u
C
G>
:J
C'"
e
LL
59.9 " New thermal governor
modeling - ggov1

59.85
...
" May 18th 2001 NW 1250 MWTrip

59.8 ~------=-------'------L ---L- ...l-_ _------l


o 5 10 15 20 25 30
Time in sees

Figure 2-29. Simulations with the New Thermal (ggovl) Governor Model compared with
May 18th System Test Recordings for the NW 1,250 MW trip, with AGe Switched Off.
The existing (incorrect) modeling assumes that 100% of governors respond in
accordance with its 5% speed droop governor characteristic. The new model
correctly models 'base'-Ioaded and 'load-controlled' units.

2.7.2 New Thermal Governor Modeling Approach


The New Thermal Governor Modeling effort followed a 3-step process of
Development. Validation and Verification.
"Development" of the modeling was based on the recorded responses of
the system and the recorded SCADA and disturbance monitoring
responses of the individual generating units during the 1,250 MW
Northwest Trip Test of May 18, 2001 .
"Validation" of the model was performed by simulations and comparison
to recorded responses of the 750 MW Hoover trip test on May 18,2001. An
additional validation was with data from a previously conducted trip test
on June 7, 2000 when 750 MW was tripped at Grand Coulee. All the three
staged tests were performed with AGCs switched off so that the responses
were a result of governor pick up and load response only.
Additional "Verification" of the model was performed by comparing
simulations with the new model with recordings of several recent
generation trips in the WECC ranging from 950 MW to 2,800 MW.

2 - 51
2.7.2.1 Analysis of Test Data - Thermal vs Hydro Units
Test data and recordings of the May 18,2001 test, from both SCADA and
disturbance monitors, were collected from the control areas and generator
owners. This included generator responses and system recordings of frequency,
voltages and flows at critical 500 kV buses and interties. The analysis of the data
showed that most of the hydro units were largely very responsive to frequency
deviations and it was concluded therefore that the un-responsiveness was due
mainly to the thermal units . A map of the location of hydro versus thermal
generation in the Western Interconnection is shown in Figure 2-30.

Generation
From June 2000 WSCC
l.-""'Cl & R......,..",.. SunTn:>ry
Total Capacity = 158,501 MIll
As Of Jan. 1, 2000


_Coal
_ f-tjdro
Southern
ClIlfornia
AZ
I
Gas

_ Other
Nudear
V\l nd

Figure 2-30. The Location of Hydro and Thermal Generation in the WECC

The analysis of test data was a huge effort because the total thermal generation
on-line during the May 18th test was about 67,000 MW out o f a total WECC
generation of 91,000 MW . A power flow base case was created to specifically
model system conditions during the tests.
The scrutiny of over 200 SCADA response recordings of generator electrical
power indi c a te d a characteristic pattern of responses . The units appeared to
have responses that could be classified in the following three broad categories:
1. 'Base ' loaded units (typically units under 'limit' control, or for gas turbines,
under temperature control) that showed no sustained MW pickup, or
response, following a frequency dip in the system.

2 - 52
2. A MW response that appeared to be an initial temporary pickup of
generation but which decayed with time. This was further classified as a
'fast' decay response (Code T1) or a 'slow ' decay (Code T2) response.
These are typically units with MW load (power) controllers. These codes
also represented units with thermodynamic responses that could be
typically modeled as load (power) controllers.
3. A sustained pickup that did not appreciably decay with time (Code T3) .
Each of the 1,100 thermal governor units that were analyzed was given a code.
About 60% of th is thermal generation was 'base' loaded. Typical responses for
units that were coded T1 to T3 are shown in Figure 2-31 depicting the generator
"responsiveness " in varying degrees. steam thermal units were coded T1 to T3,
and corresponding gas turbine units coded Gland G2 for fast and slow
controller responses.

SCADA Plots of Units - May 18th 2001 Test with AGC ott- NW Trip 1250 MW
430 74 t---""'""""::::--~-~---t
CODET1 CODET1
425
72 1--_ -1
420

415 70~-----~-~
-50 o 50 100 150 -50 o 50 100 150
485 490 t---~-~-~---t
~
~
... 480 CODET2
485
.....
0
...
III
4)
475
l: 480
4) 470 CODE T2
(!)
465
-50 0 50 100 150
760

740

720 ol---_-_-~----4
-50 o
CODET3
158

156
J:ETI
0 50 100 150

Figure 2-31. SCADA Recordings of Typical Thermal Units, Coded 11, T2 and T3
to Denote Fast, Slow and Sustained Governor Responses respectively during the May 18th
1,250 MW Trip Test (for which the system frequency response is shown in Figure 2-27)

Where SCADA data was not available for a specific unit. information obtained
from a survey of owners/control areas, regarding the base loading or
responsiveness of their units was utilized in the selection of the turbine-governor
code.
A simple block diagram of the thermal plant governor and controls is shown in
Figure 2-32 (from Reference [47]). The governor is a Proportional-Integral-
Derivative (PID) type with the classic permanent droop feedback, typically 5%.
The turbine is represented by a typical lag-lead transfer function. In the new

2 - 53
governor model, "base" load operation is simulated by setting the limiters to limit
the turbine power to a preset value. An additional MW power (load) controller is
included to model Code T1-T3 and Code G 1-G3 units. This is a simple reset
controller with its gain (Kimw) typically having values of 0.01- 0.02 per unit for
"fast" controllers, 0.001-0.005 per unit for "slow " controllers, and 0 for no load
controller action (Le. a fully responsive 5% droop unit) . The detailed block
diagram of the thermal governor model (ggov 1) is shown in Figure 2-33. (See
Reference [6] for further details of the model and the parameters.)

Freq uency

Elect rica l Po we r

Figure 2-32. Block Diagram of the New Thermal Turbine-Governor


showing 'Base' Load/Limiter, and MW Load Controller Features

2 - 54
if Dm >0 spM d

Ldref
( Ldre f,ll<t ur b )+win l 1 - - - - - - - ,
Pmech

speed

qcvervcr au pu
,---,-- ..:.;
valve st roke
rselecl
1 - electrical pow er
- 1 - valve st ro ke Flag
- 2 - govemor cutout 1 - fuel flail' proportional io s peed
o - lsoc hronous o- fuel flow independent of ~peed

Note : T~Hl Kpgov/I<lgov on<J Kploll d/ KlIoad cont roller" Include trac king loglc t o
erlsure smooth trar,sfer betwe en active co n1rollers. This logic is not shown.

Figure 2-33. Block Diagram showing the Basic Relationships of the


Turbine-Governor Plant Model ggovl (ref. [48])
Numerous sensitivity studies were performed to determine the effect of varying
parameters in the dynamic database before the final selection of the
parameters for each unit .
The two new models developed by GE for use in WECC studies are the ggov 1
and the Idbl models. The ggovl model is a generic thermal governor/turbine
model that incorporates base loading and load controller effects and was
described in [6]. The load controller model, Idb 1, is similar in structure to the load
controller portion of the ggov 1 model. Thermal units with load controllers may be
represented by the ggov 1 model, or by an older model, such as the ieeeg 1 [47],
augmented by the Idb 1 lo a d controller model. See References [6] and [49] for
details of the two models.
Upon initialization, base-loaded units and load-controllers are assigned MW set
point values in the ggov land Idb 1 models equal to the generator dispatched
value specified in the power flow data. The output of the unit will reset to the
MW setting value of the load controller at a speed controlled by the value of
Kimw in the ggov 1 model (or KI in modelldb 1.) The major validation effort was
in selecting the correct load-controller characteristic for each unit. l.e. the gain
Kimw, so that the simulated MW response of the unit to the frequency deviation
corresponded closely to the recorded MW response during the test.

2 - 55
2.7.2.2 Response Code Classification for Selection of Thermal Turbine-Governor
Data:
The designated codes are presented in Table 2-5 (references [6] and [49]).

Table 2-5 . Principal Parameters of the New Thermal Turbine-


Governor Model ggovl for the Various Designated Codes
The principal parameters of the model are:
P I D
Co d e r Tb Tc KpgoY KigoY KdgoY Kimw
Tl Fast lo a d .05 10 2 10 2 0 00.05 to 0.02
con trolle r
T2 Slow load .05 10 2 10 2 0 0.001 to 0.003
controller
T3 No load .05 10 2 10 2 0 0
c o ntro lle r
Gl With load .05 0.5 0 10 2 0 0.01 to 0.02
c o ntro lle r
G2 No load .05 0.5 0 10 2 0 0
c ontroller

r Permanent Speed Droop, pu


Tb Turbine lag time constant, secs
Tc Turbine lead time constant, secs
Kpgov Governor proportional gain, pu
Kigov Governor integral gain, pu
Kdgov Governor derivative gain, pu
Kimw Load (power) controller gain, pu

2.7.2.3 Model Validation with May 18, 2001 Test Data


The results of the simulations for validation and verification of the 'new' model
(red plots) compared with the real time event frequency recordings from
disturbance monitors (green plots) are shown in Figures 2-34 to 2-39. Simulations
performed with the inc o rre ct existing models (blue plots) are also shown for
comparison. The existing (incorrect) modeling assumes that 100% of governors
respond in accordance with its 5% speed droop governor characteristic. The
corresponding SCADA, or disturbance monitoring unit, responses are shown in
green .
Simulations of the generator electrical power of two typical thermal units with a
'fast' and a 'slow ' load controller respectively during the May 18, 2001 NW Trip
Test are shown in Figure 2-33 and 2-34. The resolution of SCADA at 4 sec intervals
does not pick up the detailed electrical power swings as seen in the simulations,
but it does show the overall general response as an envelope (green plot) .

2 - 56
Figure 2-35 shows the simulation of a typical base-loaded unit. The characteristic
initial peak typically seen at the start of the response is inertial. (With 4 second or
longer SCADA responses, this initial peak is seldom captured .)
With the new thermal turbine-governor modeling included for about 1100
governor models, the accuracy of the new modeling WECC system frequency
response is evident in Figures 2-36 and 2-38 for the May 18,2001 trip tests.

CRAIG 2: May 18th 2001 Test NW 1250 MW Trip


430

_______SCADA recording - May 18 2001 test


....
428
\ ....-- Base case (existing modelin g)

426 New thermal governor ggov1


:s: simulation
...
:2
kim'o,N=0.005,420 MW setting
....l'Il0 424 for load controller
...
G>
l:
G>
e 422

420

418

416
-20 o 20 40 60 80 100
Time in secs

Figure 2-34. Simulations with the New Turbine-Governor Model of a Thermal Unit
with a 'Fa st' Load Controller of Code 11 (red plot) compared with
its May 18th Test SCADA Recordings (green plot).

2 - 57
Slow Load Controller Unit: May 18th 2001 System Test - Tripped 1250 MW in NW
485 ,---------,- ---,----------,,---------,- - - ---,

May 18th 2001 SCADA recording


480 of NW 1250 MW Trip
"---~

...o
... 475
1U
Q)
cQ)
o
New thermal governor ~
470 ggo v1 simu latio n ----

Note : kimw=0.0015 for the unit load controller


465 L - -I- ---.JL- ....L ---.J --'

o 20 40 60 80 100
Time in secs

Figure 2-35. Simulations of a Thermal Unit with a 'Slow' Load Controller of Code 12
compared with its May 18th Test SCADA Recordings

PaloVerde Gen: May 18th 2001 System Test Tripped 1250 MW in NW


1290 ,------,------r-- - - , - - - - - - - - , - - - - - - - . - - - - - . - - - - - ,

1285
New thermal governor ggov1 modeling

1280
:s:
:2

~... 1275
Q)
cQ)
o
1270 -

1265 SCADA recording

1260
0 5 10 15 20 25 30
Time in sees

Figure 2-36. Illustrating the Simulation of a Typical "Base Loaded" Unit


compared with its May 18th Test SCADA Recording

2 - 58
May 18 2001 Test NW Trip 1250 MW Malin 500 kV Bus Frequency

60

/ B ase case (existing models)


~
I 59.95
>;
u
l:
Q)
::J
0-
e
11.
59.9 " New thermal governor
modellnq - ggov1

59.85 May 18th 2001 NW 1250 MW Trip

59.8 '-----_ _ ---'- -'-- ...L- l . . -_ _ ----.l ---l

o 5 10 15 20 25 30
Time in sees

Figure 2-37. System Frequency Response Simulations with the New


Governor model (red plot) compared with May 18th Test Recordings
(green plot) for the NW 1,250 MW trip, all AGCs Switched Off

May 18 2001 Test Hoover Trip 750 MW . Malin Frequency


60.02 r - - - - - - r -- - - , - - - -. , . - - - - r - - - - - - , . - - - - ,

60
N
I
Ill-
::J
CO
> 59.98 'Base case (existing models
ex
o
o
II) New thermal governor
~ 59.96 modellnq- ggov1
III
:2 \
n;
>.
~ 59.94
Q)
::J
0-
Q)

U:: 59.92
\ May 18th 2001 SW 750 MW Trip

59.9 '-----_ _ -----'- -'-- ...L- l . . -_ _ --I ---l


o 5 10 15 20 25 30
Time in sees

Figure 2-38. Governor Model Validation - Hoover Dam May 18th Test Simulation,
750 MW Generation Trip, all AGCs Switched Off

2 - 59
.June 7. 200 1 Test750 fo,fPlV Cou lee Trip :Malin Frequency
60 .02 r - - -- , - -----,- - --,-- - - - - , - - - - , - - - - - - ,

60 l

..Bas e case (exis ting modeling)


59 .98
N
::x:
;:;:. gg ov1 simulat!on
.... -,
~ 59.96
;,
c:r
e
LL
J une 7th 2000 Test recording
59 .94 .

59.92

59.9 '- - - - - - 1 L I I

o 5 10 15 20 25 30
Time in sees

Figure 2-39. Governor Model Validation - June 7, 2000 Test Simulation,


for the 750 MW Grand Coulee Generation Trip, All AGCs Switched Off

2.7.2.4 Model Verification with Random System Trip Data


Figures 2-40 and 2-41 show simulations of two typical random large system
disturbances performed for the " Verific a tio n" of the new model (red plots)
comparing w ith d isturbance monitoring frequency recordings (green plots) . The
reason that the new model simulation in Figure 2-41 differs from the disturbance
record ing in the 20 to 30 second range is that units picking up on AGe were not
properly modeled . Many more system d isturbances were also verified that are
not shown in this section. Note that it is req uire d to use a power flow base case
that represents closely the system existing conditions during the event in order to
get a verifiable simulation .

2 - 60
August 1,2001 - Colstrip 2000 MW Generation Trip :Frequeney

60

Base case (existing modeling


59.95

59.9 Colstrip Aug.1 reeordin


N
I
~9.85
cQ)
::J

....g- 59.8
New th ermal governor ggov1 modeling
LL.

59.75

59.7

59.65 '-- '---- -'-- ...1..- -'- -'

o 10 20 30 40 50
Time in sees

Figure 2-40. Governor Model Verification. 2000 MW Colstrip Trip


in Montana on August 1. 2001

June 3, 2002 Diablo 950 MW Generation Trip : Malin 500 kV Frequency


60.04 r - - - - - - , -- -- - r -- - - - , - - - - , - -- - - . - - - ---,

60.02

60

59.98
I Base case (existing modeling)

__
:I! 59.96
~ Diablo June 3 2002 Recording
o
~
:J
59.94 . .......--..;.--
6~"""""---'I

/
0"

I.L
e 59.92 New thermal governor
ggov1 simulation
59.9

59.88

59.86

59.84 ' - - - - - - - ' - -- - - - ' - - -- - - ' -- - - --'--- - - ..1...-- ----'


o 5 10 15 20 25 30
Time in secs

Figure 2-41. Governor Model Verification - 950 MW Diablo Generation Trip


in California on June 3. 2002.

2 - 61
2.7.2.5 Sensitivity of Parameters
A number o f sensit ivity studies were performed to determine the effect of varying
governor parameters in the dynamic database. Clearly, the greatest effect was
the selection of the base loaded (non-responsive) generators, followed closely
by the choice of " fa st" or " slo w " load controllers. The method of governor code
selection was described earlier in this section. Figure 2-42 shows the effect of
varying the selection of the base loaded and load-controlled generators.

Sens itivity Studies May 18 2001 Test - NW and SW Generation Trips


60 - ..- - - - - , - - - - - - , - -- ---,------,------,--------,

59.98 ~_.--r-~--- .
~/<' > Base case simulations (existing models}"", NW
-~ / -sw-
N
I 59.96 l~ "_.~._~ ......~ .. - - - - tItiII'_,~-=-::.':::::::::
III
:J
~ 59.94
~

o
~ 59.92
c . . "' .
ro :::S~nsiti vity range~;..:r
::?: 59.9
'til l :"" new thermal governor
.... si mu lations
g>- 59.88 "
.
I-
Q)
:J

e 59.86
0-

L1. ~'~" ~aY1 8 2001 NW 1250 MWTrip record ing


59.84

59.82 L . -_ _- - ' - ----'- ----'--- -'- --'--_ _-----'

o 5 10 15 20 25 30
Time in seconds
Figure 2-42. Effect on System Response of Varying 'Base' Loaded
and Lo a d Controlled' Unit Detections.

Sensitivity studies showing the effect o f fast and slow load controllers on t he
system are demonstrated by a "macro" study varying the speed of all load
controllers in t he system, see Figure 2-43 . Kimw is the gain of the load controller -
se e Figure 2-33 and Table 2-5 for details.
The e ffect of varying Kimw on the response of a specific unit, and therefore on a
Imacro scale on the frequency of the entire system, is illustra te d in Fig ure 2-43,
I

varying from a quick-acting controller (Kimw = 0.01) to a v ery slow controller


(Kimw = 0.00 1). The final selection for the developmental database was
evaluated from indi vid ua l unit SCADA responses. The faster the load-controller,
the quicker its generator MW response re t urns to its set-point, and t he closer its
response is to a ' b a se-lo a d e d' unit. Conversely, the slower the load-controller
response, the longer its MW p ickup is sustained to a frequency deviation event.

2 - 62
other sensitivities studied included varying the proportional and integral gains of
the PID governors (the derivative gain was maintained at zero) and varying the
Tb. Tc parameters of the turbine model. These studies resulted in varying levels of
impacts, but generally less than the effect of base loading, or varying the load
controller gain, Kimw, of the thermal units .

Diablo 950 MW Trip - Malin Frequency- kimw sensitivity

N
::r:
~ 60
I1l
>
..::.::
o
o
It)
kimw= 0.001
.559.95
Ri
:2
1U
>.
u
~ 59.9
:::J
0"
Q)
U.

59.85 L -_ _--'--_ _- - ' -_ _----'- -'------_ _---l...-_ _- - '

o 5 10 15 20 25 30
Time in secs
Figure 2-43. Effect on System Response of Varying the Load Controller Gain Kimw
on a 'Macro' Basis. Kimw of 0.01 is for a 'fast' controller and 0.001
is for a slow controller operation.

2.7.3 WECC Approval of the ModeJ. Populating the Database with Generator
Owner's Data
The new governor model went through an intensive approval process in the
WECC, including numerous presentations at the Operation and Planning
Committees, and various subcommittees and workgroups, before it was finally
approved . It was clear that the initial 'developmental' database that was
created could not, and should not, be used for real-time operation studies or to
set limits for intertie operation due to risks and liabilities. An intensive and
coordinated effort was therefore launched in the WECC to obtain 'validated'
governor model data from the generator owners to replace the
'developmental' data created for validation studies of the New Thermal
Governor Model. This effort included a WECC Workshop, issue of Guidelines for
Selecting and Validating New Governor Models, and issue of new techniques for
Model Validation and Methodologies for assisting in the process of selecting

2 - 63
model parameters and validating it. The timeline for the data effort was driven
by the need to include the new governor models in time for performing critical
operating studies for the 2003 summer season.
To assist in the selection of the appropriate model, and the governor parameters,
the generator owners were encouraged to answer typical questions with
reference to the unit response diagram in Figure 2-44 to describe the response
that best characterized their unit's electrical power response as recorded by
disturbance recorders or SCADA.

Generator J'vIW 30 sees


Respons
1 - - - - - - -- - - - - -- - - - - + - - - - - ----, c
Responsive - Code T3 - 0 controller

~
Controtle, - _
Fast
Controller - C d~C CodeT2 ~

c Base Loaded

o sees Time in seconds 30 sees

Figure 2-44. Unit Electrical Power Response Diagram and Code Classification
(see reference 2). The initial electrical response 'AB' is 'inertial' and is common for all
responses. The responses BC will end up in one of 4 'boxes' characterizing '
base-loaded', fast or slow controller, or 'responsive' operation of the unit

Typical questions to be asked by the Owner before selection of the appropriate


model Code w ith respect to the Response Diagram in Figure 2-44 (reference [49])
are:
1. Is the generator unit normally operated in a mode that can be
considered base loaded? (For definition of base loaded, see lowest
Base Loading Response Box 25.)
2. Is the generator unit normally operated under load set point control, or
any other mode of controls- that will override automatic action of the
governor responding to changes in system frequency?
3. If the answer to question 2 is yes, is the response time of the dominant
controller fast or slow as indicated on the time scale in Figure 2-44. (See

The initial electrical response' AS' in Fig.18 is 'ine rtia l' and is common for all responses.
25
The temperature limiter in a gas turbine is an example of a limiting control of the unit
26

output.

2 - 64
Response Boxes for 'Fa st' Controllers Code T1 and 'Slow' Code T2
Controllers.)
4. Is the generator unit normally operated in a mode that can be
considered Responsive? (See Upper 'Responsive' Box for Code T3.)
5. Does the generating unit normally respond to AGC signals?
It is understood that units may be operated in different modes from day to day,
or even hour to hour, and that the responses to these questions will vary
accordingly. In these cases , it is up to the owner to decide which mode the unit
is most likely to be operated in, at any given time, keeping in mind that most of
the power flow base cases of concern are intended to represent the system
during seasonal peak loading conditions. See Reference [49] for details of the
model validation technique . Figure 2-45 shows an example of model validation
for a specific unit.

May 18th 2001 NW Test - 1250 MW Tr ip - Comparing slow and fast contro llers
460 +-_ ---L._ _ ...L..-_---'-_ _....L...-_ --'"_ _........._ ---'_ _........._--+

450

440

:s: 430
:2
....
~.... 420 -r-"-~d~!!!~'
Q)
r:::
Q) ggov1 fast con t roller
(!) 410

400

390

380 +----r---r--~---r----,.---r-----,---r----+
o 20 40 ~ 00 100 1W 1~ 1~ 100
Time in secs
Figure 2-45. Electrical MW Power Simulations of the New Turbine-Governor Model
(ggov1) comparing 'slow' and 'fast' load Controllers using a Small Equivalent System and
comparison with its May 18th Test SCADA MW Recording. The 'slow' load controller is
clearly the correct choice of model for this unit.

2 - 65
To assist in the governor model selection and validation, WECC made available
to generator owners the system frequency recordings of several past
disturbances in addition to the May 18,2001 tests (750 MW Hoover trip and the
1,250 MW NW trip). In the future, when a large generation trip occurs that is
suitable for model validation, it is the intent of WECC to send out a notification
within 24 hours so that the generation owners can retrieve the captured
validation data. Each owner would also record the manner in which his/her unit
was operated at that time. A file containing the system frequency versus time
data to be used for validation is also planned to be sent out by WECC. The
owner may also request SCADA records for his/her unit from the Control Area
Operator. These processes are currently being revised in keeping with new
confidentiality or security rules for data transmittal. But the general intent is the
same - l.e. to assist the generator owner to select and validate his governor
model.
The alternative to SCADA data is for the generator owner to install disturbance
monitoring equipment to capture data during random system disturbances.
Since the data required is system frequency, power and voltage, versus time, an
8-channel Data Acquisition System (DAQ) with 3 voltage signals and 3 current
signals would normally be adequate. Signal processing software developed by
BPA for point-of-wave recording for governor responses has been made
available to owners who need to use this method of data retrieval. The cost of
the entire monitoring equipment, including DAQ card, modules, chassis and
cables is reasonably low.
2.7.3.1 Some Governor Modeling Issues
Operation of Units Differently at Different Times:
While there are many SCADA recordings of units showing a consistent method of
operation, there are also cases where units are clearly being operated differently
at different times. A typical case is shown in Figure 2-46. This figure shows
recordings of the operation of the same unit for different events, indicating
operation as a base loaded unit during some events, and with a load controller
in others. For such units, the owner clearly has a key decision to make, when
submitting governor model data, whether the unit would be designated in one
or the other operation mode. This is a decision that only the owner can make for
his unit. This example also illustrates that it is correct for the control area to
assume arbitrary model data for a particular unit without the agreement of the
owner. Operating a unit differently at different times makes its response
unpredictable, thereby also making its simulation and impact on the system
unpredictable.

2 - 66
Responses of the same unit during different events
1.07 +--------'--_ _.....L... .l....- --1. --I-
ote: The generator initial power is shown to identify each recording
01
c
E 1.06
Q)
(II

~ 1.05
jij
E 343MW
t:
~ 1.04
o
.....
2
: 1.03
Q)
a.
t:
";: 1.02
~
o
0...
.... 1.01
o
1U
~
Q)
1 L.J~~~~~i
o
0 .99 +-------,------.-----..---------r-----+
o 50 100 150 200 250
Time in secs

Figure 2-46. Generator Electrical Power SCADA Recordings of the Same Unit for Different
Events. Shown on a per unit scale where 1 pu equals the initial loading of the unit (varies
between 343 and 470 MW). The unit was clearly operated as a base loaded unit during
some events, and with a load controller, at other times.

Effects of AGe:
For studies extending to long periods, such as for system oscillations and dynamic
voltage stability, it is desirable to model Automatic Generation Control (AGC).
Comparison of the system recordings of the May 18th Test when all AGCs were
switched off, and the system recording of a random trip of the Colstrip 2,000 MW
plant on Aug.1, 2000, clearly indicates that AGC does make a difference in the
frequency response of the system . This is illustra te d in Figure 2-47.

2 - 67
Comparing August 1, 2001 - Colstrip 2000 MW Trip
and May 1 2001 NW 1250 MW Trip Test
N
60
I

,.
(I)
::J 59.95 ,May 18 2001 NW 1250 MW Trip Test without AG
co
>~
o 59 .9
o
LD
C
.~ 59.85
:2
rti --- Colstrip 2000 MW trip
>. 59.8
u Aug.1 2000 disturbance recording
cQ)
5- 59.75
Q)
10-
LL
59.7

59.65 '-----------'--------'-----------~
o 50 100 150
Time in secs

Figure 2-47. Disturbance Monitoring Recordings comparing Two


Real Time Recordings - with and without AGe.

Owners whose units are on AGC for a substantial part of the time of interest (for
example, peak summer operation case) could, in consultation with their control
area operator, designate their units as AGC units for the purposes of modeling.
These units would be temporarily designated as Code T3 or "responsive" units
until more accurate AGC modeling has been established.
Proper AGC modeling is an ongoing task in the WECC . AGC could start w ithin 10
to 15 seconds (or even earlier in some control areas) following loss of generation
in the area, change in freq uency (frequency-bias operation) , or tie line flows.
Practices and systems vary in the different control areas in the WECC. It is
recognized that since the new thermal governor modeling approach represents
unresp o nsive units more accurately, the effects of AGC should be appropriately
represented for more accurate simulations of disturbances.

2 - 68
2.7.4 Improved Hydro Plant Responses
An important finding of the impact of the thermal "unresponsiveness" on
modeling was the greater importance of the "responsiveness" of hydro
modeling, and hence the greater demands on more accurate hydro modeling
(see reference [50]). The improved modeling of thermal plant response that
results in a lower pickup of thermal plants also results in a corresponding increase
of pickup by frequency 'resp o nsive ' hydro plants. Figure 2-48 shows the greater
pickup of a typical hydro generator in the May 18th Test simulation . Because
thermal plants in the WECC are predominantly located in the South, and hydro
generation is predominantly in the Northwest (see Figure 2-34), the improved
simulation of unit MW pickup and power flows across the system, particularly in
intertie flows between the Northwest and the South, is critically significant in
operation and planning studies.

May 18 2001 Test NW Trip 1250 MW Hoover Gen Power


108.5 , - -- - - - , - - - - - - - - . - - - - , - - - - - - , - - - - --,---------,

108

107.5
:s:2
...-
~ 107
o
n,
...o
1U
... 106.5
Q)
~Base case (existing modeling)
cQ)
o 106

105.5

105 L - -_ _- . l . . --'-- ----'--- -'- ...l...-_ _- - - - '

o 5 10 15 20 25 30
Time in seconds
Figure 2-48. Showing Hydro Plant Responses: Improving the Accuracy of the New
Thermal Governor Modeling {ggovl} increases the Generator Pickup
of a Typical Frequency 'responsive' Hydro Unit {red plot}.

2.7.5 System Simulation Impacts of the New Thermal Governor Modeling


It is important to realize that the correct modeling of the governing
responsiveness of units resulted in significant system simulation improvements in
the WECC in several areas. The following are some of the important impacts of

2 - 69
the new thermal governor modeling on major system operation and planning
areas:
System frequency responses can be predicted more accurately for
large generation trips.
Improved modeling of hydro versus thermal generation responses.
Improved under-frequency and load shedding studies, involving large
generation trips and/or system islanding
A more accurate prediction of critical intertie flows and dynamic limits
is obtained for operation in a system such as WECC where responsive
hydro generation is located in the north and largely thermal
generation is in the south.
Improved assessment of system oscillations and damping, because of
the improvement in the simulated redistribution of power over interties.
More accurate assessment of Frequency Responsive Reserves (FRR)
and Spinning Reserves.
More accurate post-transient ('governor') power flow studies involving
large generation trips.
2.7.6 Summary and Conclusions
A new thermal turbine-governor modeling approach, based on improved
simulation of base-loaded units and load-controlled units, has been developed
in the WECC. The development of this model went through an extensive study
process that included validation to staged WECC system tests and verification
with respect to numerous large system disturbances. After approval of the
model for use in WECC, an intensive and coordinated effort was launched in the
WECC to obtain validated governor model data from the generator owners. This
effort included a WECC Workshop, issue of Guidelines for Selecting and
Validating New Governor Models, and issue of new techniques for Model
Validation and Methodologies for assisting in the process of selecting model
parameters and validating it. The new governor modeling approach is being
currently used in all operation and planning studies in the WECC.
Further work is proceeding in several areas including improved generator model
data, AGC modeling, studies of system oscillations, and post-transient power flow
modeling.
While the interest as described in this section was specifically to governing
relating to the WECC, the Western Interconnection in North America, the general
principles of the new thermal governor modeling approach clearly apply to all
interconnections, large and small.

2 -70
2.8 New York Observations of Generator Governing Response
New York Independent System Operator (NYISO) has installed a specialized data
acquisition system capable of continuously sampling and storing telemetered
analog data at a rate of 10 samples per second. The data sampled includes bus
voltages, line flows, and real power outputs of large generators. With the
availability of this data, it is possible to examine the response of some generators
in the New York control area to significant frequency excursions. The high
sampling rate allows estimates of the governor droops to be made for
generators which exhibit governing response.
Several specific incidents of frequency excursions in the US Eastern
Interconnection are presented here. These incidents were monitored in New
York and were produced by a sudden loss of generation or loss of load. The
events presented here were selected to present good examples of the kinds of
unit and system responses; they are necessarily a small subset of the many
thousands of events. Also, generation is identified by class rather than name of
facility. Plots of the droop characteristics were generated for each of the units
studied by plotting the change in system frequency versus the MW change of
the unit over a 100 second period.
2.8.1 Loss of Cook Units, April 2003
On April 24, 2003, both Cook units, totaling 2,120 MW, tripped approximately 30
seconds apart. The time of the first unit trip was 03:28:19 EDT. The frequency
change for this event was 82 mHz; thus, the observed Beta for this incident was
25.8 MW/mHz.
For this incident we have included observations of the MW output of seven units,
as shown in Figure 2-49. Droop characteristics of some of these units are plotted
separately in Figures 2-50 through 2-52.

2 - 71
;J
System Frequency - Loss o f Generation #1

i ;5985:~ ~
~ . -
o 20 40 60 80 100
Gove rnor Response - Hydro Unrt #1
105

~ 1 I I : 1
~ 0 9~ i-"~~"""""""'""'"'''' ... .,,"'' '. .,]
o 20 40 60 80 100
Governor Re sponse - Hydro UCIt #4

o 20 40 60 80 100

o 20 40 60 80 100
Governor Responsf - Tho?rmal Urnt #4

o 20 40 60 80 100

: :E ::'- J
Gov ernor Response - Thermal Urut #6

3
,.
'"
~.

~
0
c,

0 20 40 60 80 100
Gov ernor Response - Therm al Unit #9

3
2
a; ~
~
0
c, 0 [

06:
0 20 40 60 80 100
Go vernor Response - Thermal Unit #11

20 40 60 80 100
Time Is )

Figure 2-49 : MW Response of Units following the Loss of both Cook Units

2 -72
Governor Response - Hydro Unit #1
0.01 , ; T f r r. ............................

a
-0.01

-002
N
~ - 0.03
()
a3 - 0.04
::J
0-
Q)
'-
LL
<J

-0.07

- 0.08

-0 .09 '------'------'-----'-----'-------'---~
-0.02 o 0.02 0.04 0.06
~ Power (pu)

Figure 2-50: Droop Characteristic following the Loss of both Cook Units

Gove rnor Response - Ther mal Unit #6


0.01
... ~

-0.01

-0.02
N
I 03
>.
o
c:
Q) - 0.04
::J
0-
~
LL - 0.05
<J
-0.06

-0.07

- 0.08

- 0.09
-0.02 o 0.02 0.04 0.06
~ Powe r (pu)

Figure 2-51 : Droop Characteristic following the Loss of both Cook Units

2 - 73
Governor Response - Therma l Unit #9

..~

-0.01 i o

'( ,
..

I
N
- 0.02 1
:
.1-
?
..1

~ -0.03 i
>.
g I
:
!
Q) -0.04 ;' .J
::J : !
0-
Q)
:
: I
u:: -0.05 f
I
<1 -0.06 l.
,
-O.OlL -~

-0.08 :' .,,

-0.09 1'-- -'"- - - - -l-- - - - - - - - -L- -L..-_-'


-0.02 o 0.02 0.04 0.06
~ Power (pu)

Figure 2-52: Droop Characteristic following the Loss of both Cook Units

All of the hydro units showed a response to the frequency drop. Hydro unit # 1
showed a response consistent w ith a droop of approximately 10%, while hydro
un it #2 responded with a droop around 5% during the frequency transient but
then appeared to reduce output afterward, even though little frequency
recovery had occurred. It is interesting to note that hydro unit #4 , which was in a
pumping mode at the time, also showed a frequency response; it is not clear if
this is predominantly due to governor action or a frequency-dependent
characteristic of the pumping load.
Both nuclear units showed little if any MW response to the frequency change.
The inertial response of the units is evident in the data, but the MW output was
o therwise virtually unchanged.
The thermal units showed several different types of responses. Unit #6 is
indicative of an active turbine-governor with a 7.5% droop. It is noteworthy that
this unit was running well below its maximum. Units #5 and # 10 showed no
response (other than inertial), much like the nuclear units . Unit #4 appeared to
show a slight positive response at the times of the frequency drops, but shortly
afterward the real power output of the unit dropped noticeably, becoming less
than the initial value prior to the frequency change. Unit # 11 output dropped
before the first unit trip but increased some at the time of the second un it trip.
Finally, unit #9 showed a significant response to the second frequency drop but
then began to reduce its output to pre-event levels . This is an indication of an

2 -74
outer-loop MW control with a longer time constant that eventually counteracts
turbine-governor action [6], [16].
2.8.2 Trip of Raccoon Mountain Pumped storage Plant, May 2004
On May 31 , 2004, the entire Raccoon Mountain pumped storage plant tripped
with all units in pumping mode. The total pumping load at the time was 1,600
MW . Such a large loss of load all at one time is extremely rare. This event
provided an opportunity to observe governor performance for a large frequency
change upward, in this case 55 mHz. The observed Beta for this incident was
29.1 MW /mHz. The time of the incident was 03:11 :49 EDT.
For this incident we have included observations of the MW output of four units in
Figure 2-53. A plot of the droop characteristic for one generator from this event
is shown in Figure 2-54.

System Frequency - Loss of Pumped Storage

~~ 6005
;;;)
60 1 [ ; C , , ~ ]
~ 60
LL 59.95 _ _ --'-_ _ --'-_ _-----'- _
o 20 40 60 80 100

o: F;;;;".:-~=d
Governo r Response - Hydro Unit #1

f o

f ~: :E':- " .~. _~


20 40 60 80 100

Governor Response - Hydro Unit #4

o 20 40 60 80 100

Governor Response - Nuclear Unit #1

lo::~ o 20 40 60 80 100

lo',:E : -: : j
Governor Response - Thermal Unit #10

0 .1
o 20 40 60 80 100
Time (s)

Figure 2-53: MW Response of Units following the Loss of all Raccoon Mountain Units

2 - 75
Govern or Respo nse - Hydro Unit #1
r . . . . . . . .. r T ..... w

I ,
0.07 l ...i
~
,,
0.06 l
N
I 0.05 1
j
,
0.03 f

0.02 '

0.01 -

01

-0.04 - 0.02 a 0.02


~ Power (pu)

Figure 2-54: Droop Characteristic following the Loss of all Raccoon Mountain units

Hydro unit # 1 showed a response consistent with a 5% governor droop. Hydro


unit #4, which was pumping, showed a response indicative of a 15% droop over
a more sustained period . Hydro unit #3 also showed a slight response. As
before, it is not known if the frequency response of the pumped storage units is
predominantly a result of governor action or a frequency-dependent property of
the load.
The inertial response of the nuclear units to the event is very evident in the data,
but the units otherwise showed no response to the frequency change. Similarly,
thermal units #4 and # 10 showed an inertial response but no indication of any
governor action .
2.8.3 Loss of Cumberland 2, November 2006
On November 9,2006, the loss of a single unit, Cumberland 2, at 1,260 MW was
observed. In this case the observed frequency decline was just over 50 mHz. This
observed Beta was thus 25.2 MW /mHz. The frequency drop began at 13:24:58
EDT.
For this incident we have included observations of the MW output of two units in
Figure 2-55 and a correspond ing droop characteristic of one unit in Figure 2-56.
Hydro unit # 1 initially reduced output when the frequency dropped, but then the
output began rising toward a level consistent with a 5% droop. This initial
reduction in output is a known characteristic of hydro plant operation and is due

2 - 76
to water inertia in units with long penstocks. Like many nuclear and thermal unit
responses for the previous two events, thermal unit # 12 showed virtually no
response for this event, other than a brief increase due to inertia .
2.8.4 Summary and Conclusions
Based on the work summarized in this section, some general observations about
the governing response provided by different unit types may be made:
1. Nuclear units, generally speaking, exhibit little or no governing
response.
2. Large hydro plants generally exhibit a significant response in the
correct direction. This may be extendable to all hydro, however smaller
plants are not as readily observable and were not included in the
analysis.
3. Fossil fired steam turbine plants, either conventional steam or
combined cycle , exhibit a variety of responses, including the lack of
any response, response in the correct direction, response in the wrong
direction, and brief responses, followed by return to pre-event levels.
These observations are consistent with observations in other regions and
interconnections, as discussed in the previous sections of this report.

System Frequency - Loss of Generali on #2

N
E. 60
:>.
'-'
c 59.95
'1'
~
0-
~ 59 .9
LL

59.85
0 20 40 60 80 100

Govern or Re sponse - Hydro Unit #1


1.05

3.
....0>
S
0
Q.

0.95
0 20 40 60 80 100

Governor Response - Thermal Unit #12


0.95

3
.3
.... 0 .9
0>
S
0
Q.

0.85
0 20 40 60 80 100
Time (s)

Figure 2-55: MW Response of Units following the Loss of Cumberland 2.

2 - 77
Governor Response - Hyd ro Unit #1
0.01 , , --..- -..- ., -..- -..-- ----,-.--.-.---..----,--.-..- -., - -,

- 0.01

-0.02
N
I

~
~ -0.04
:::J
0-
Q)

li:: -0.05
<I

-0.06

-0.07

-0.08

-0 .09 '-----'--------..:.-----~----'-- -_L-.---'


-0.02 o 0.02 0.04 0.06
tl Power (pu )

Figure 2-56: Droop Characteristic following the Loss of Cumberland 2

2.9 Summary
This chapter has p resented results from several utilities and interconnected
systems around the world, with respect to recent measurements of large
inte rc o nnect system response to generation governing . In addition, in a few of
the cases presented here (specifically US utilities) the respective utilities/reliability
councils have performed extensive studies to identify the initial disparity between
actual and simulated system response. These studies have shown some general
trends and observations. The most important are:
Models in the simulation database do not necessarily reflect the actual
control mode of many units during an event. That is, the model may
represent a fully responsive governor, while the unit may actually be
under outer-loop supervisory megawatt control.
Not taking into consideration, adequately the damping effects of
lo a d s and the effect of fault impedance on the init ial frequency
transient.
Not taking into consideration the fact that many turbine-governors
may be non-responsive, being either on temperature limit (gas
turbines) or at base load .
In the next chapter the overall conclusions of the Task Force as presented,
together with recommendations for future work.

2 - 78
3 CONCLUSIONS AND RECOMMENDATIONS
This chapter is organized into two sections. The first section reports on the
conclusions and findings of the Task Force based on material presented in this
report and discussed during Task Force meetings. The second section presents
the recommendations of the Task Force in addressing some of the outstanding
issues reported in our conclusions.

3.1 Conclusions on Present Practice


The conclusions here are not listed in any particular order of merit.

1. Good governing capability is needed on all units that are part of an


electrical island formed by a system breakup. Since the formation of
islands is rare and unpredictable, both in generation-load imbalance and
geographic extent for most of the interconnection, this requirement
provides impetus to the approach of a reward based method of
monitoring and providing financial incentive for governing response. This
observation is the root of the purpose of the Task Force, to address the
conflicting pulls of lowest possible cost of electricity without risking the
costs of a system blackout.
2. Work on the North American interconnections indicates that some of the
present modeling practices may provide inaccurate results when used to
estimate future operation.
a. Actual levels of Primary Governing Frequency Response (PGFR) as
observed on the North American interconnections are significantly
lower than expected based on past practice for estimating PGFR.
b. Methods to improve the estimation of PGFR have been
investigated and indicate that a number of factors have
contributed to the overestimation of PGFR as observed in dynamic
simulations using the model.
c. The methods identified to improve the models have been
implemented on some interconnections with resulting improvement
of the match between models and actual experience.
3. For those parts of the US-Canadian Interconnections that are at low risk of
unplanned islanded operation, the present governing response is widely
accepted by expert opinion as appropriate for maintaining security and
avoiding the costs of dispatching generation units to maintain a
governing margin on all units.
a) For the Eastern Interconnection, the total system inertia is large enough
and the governing response is great enough to avoid under-frequency
load shedding for typical disturbances, such as the loss of a
combination of large units. The largest frequency excursion in recent
decades occurred on August 14, 2003. This event involved the

3- 1
complete islanding and subsequent blackout of a sizable portion of
the Interconnection covering all or part of eight states and one
Canadian province. Despite this, the frequency deviation on the
remaining part of the Eastern Interconnection was limited to about 300
mHz (increase due to generation surplus in the intact regions). With
the exception of this extraordinary event, the largest loss of generation
events have in no case resulted in a frequency decline of more than
100 mHz, which is well above the point at which the first stage of
under-frequency load shedding commences.
b) As long as governing response is maintained above identified limits,
the Texas Interconnection is able to maintain frequency decline
above load shedding limits for the loss of the largest unit. Programs to
measure and correct problems identified have been implemented on
this interconnection.
c) The Western system is also able to avoid load shedding for the loss of
the single largest unit. Studies have been done to identify the disparity
between predicted and actual PGFR.
d) When inter-area oscillations are well behaved, the system governing
response is fast enough. This conclusion was reached at a Task Force
meeting, based upon the experience of the members present when
examining recorded frequency following loss of a significant portion of
the generation:
i) The time from beginning of the frequency decline to the
'bottom' is about 2-3 seconds in the Eastern Interconnection,
about 6 seconds in the WECC [51] and about 4-6 seconds on
the Texas Interconnection.
ii) The time from the 'bottom' of the decline to the 'settled out'
time at which frequency stabilizes is about 5 seconds in the
Eastern Interconnection, and about 15 seconds in the WECC
[51] and Texas Interconnection.

iii) The time from onset of the event to the time when frequency is
maintained at a new level by governing and frequency
regulating effects is in the range of 10 to 20 seconds.
Conventional North American practice is to have 'secondary
frequency regulation' or AGC begin to take effect not sooner
than 30 seconds, and possibly 60 or 90 seconds.
4. Since there is no inherent reward, but significant costs, for generation
owners and operators to provide governing, there will be a trend to
reduced governing response. This trend may have been underway for
decades [5]. If current declining trends continue, all interconnections will
eventually experience unacceptable security risk.
5. Traditional methods of insuring adequate PGFR may not work under
restructured systems because the lines of authority have been severed
and/or replaced.

3-2
6. Testing or monitoring to determine governor performance on generators,
and to determine the mode of operation of the unit (e.g. base-loaded, on
temperature limit, on outer-loop megawatt control) in order to report this
information to the appropriate transmission system planning staff is not
done in many regions on a routine basis.
7. Good governing response is most important in portions of an
interconnected system that are at higher risk of unplanned islanded
operation. Generally, these portions are areas that frequently
import/export energy for economic advantages over relatively few
transmission corridors. Thus, they are essentially peninsular type systems.
Some examples in the US-Canada include peninsular Florida, eastern or
Maritime provinces of Canada, the provinces of Manitoba, Saskatchewan
and Alberta, southeastern New York including the New York metropolitan
area and Long Island, and Southern California. An international example
is the country of Italy. Thus, good governing response, backup under-
frequency load shedding and proper behavior of the generating units
under low/high frequency events is critical for these systems. The key
factor, for these peninsular systems, is the ratio of the total megawatt
import/export to the total load.
8. PGFR cannot be assured without measuring whether or not it is being
provided and who is providing it.
9. There appears to be a noticeable decline in the total megawatt per Hertz
response of the interconnected system to the loss of generation in some
systems following the restructuring of the power industry from a vertically
integrated utility to a market-based structure [5], [53]. This is driven by a
number of factors. The two major factors appear to be:
a) Since there is no monitoring of or compensation for primary frequency
regulation in many deregulated markets, most large generating
facilities will operate in more efficient operating modes. For example,
turbine-generators may generally not carry any reserve megawatt
margin thus disabling their ability to provide primary governor response
following a disturbance that result in system frequency decline.
b) The advent and increasing deployment of new generation
technologies such as wind, solar and cogeneration facilities that
generally do not contribute to primary frequency regulation.
1O. Some general observations about the governing response provided by
different unit types can be made:
a) Nuclear units, generally speaking, exhibit little or no governing
response. This observation is consistent with results found in [38].
b) Large hydro plants exhibit a significant response in the correct
direction. This conclusion may be extendable to all kinds of hydro
units, however smaller plants are not as readily observable and were
not included in the work reported by NYISO.

3-3
c) Fossil-fuel steam turbine plants, combined-cycle units and gas turbines,
exhibit a variety of responses, including the lack of any response (e.g.
base-loaded or on temperature limit), response in the correct direction
(governor control), insufficient response in the correct direction,
response in the wrong direction (frequency dependence of maximum
output for large frequency excursions), and various brief responses
followed by return to pre-event levels usually when on outer-loop
megawatt control.
d) Wind turbine generators do not contribute to primary frequency
regulation. Furthermore, most designs do not contribute to inertial
response either. Work is being done to demonstrate how these issues
may be addressed for wind generation [21], [22].
3.1.1 Conclusions on Difficulty Observed in Computer Simulations and Modeling
Among the several generation/load imbalance disturbances that have been
analyzed after-the-fact by computer simulation, essentially all the work has
shown clear disparity between measured interconnected system governing
response. In many cases the cause(s) of the disparities have been identified and
since rectified in the model(s). Many utilities have observed this. The main causes
appear to be the following:
1. Models in the simulation database do not necessarily reflect the actual
control mode of many units during an event. That is, the model may
represent a fully responsive governor, while the unit may actually be under
outer-loop supervisory megawatt control, under sliding pressure control
etc.
2. Simple errors in the data base such as having the turbine-governor
modeled on generator MVA base and not ensuring that Pmax (maximum
turbine MW power output) properly reflects the actual turbine capability
under the present ambient conditions.
3. Not adequately taking into account the damping (governing) effects of
loads and the effect of fault impedance on the initial frequency transient.
4. Not taking into consideration the fact that many turbine-governors may
be non-responsive, being either on temperature limit (gas turbines) or at
base load or on outer-loop megawatt control (item 1 above).

3.2 Recommendations
3.2.1 Recommendations Related to the Disparity Between Simulations and
Actual Observed System Response
For the simulation and modeling issues the following recommendations are
made:
1. All interconnections should compare their modeling results to actual
experience.

3-4
2. Care should be taken in modeling governor response to capture the
behavior of common control functionalities such as:
a. Temperature limits on gas turbines
b. Base loaded units that are not responsive
c. Steam turbines operating under sliding pressure mode
d. Units under outer-loop megawatt control
The discussion in chapter 2 on the WECC approach to modeling the
response of thermal units discusses much of this [6], [1]. All these features
need to be made available in commercially available software used for
power system simulations (See recommendation 2. under section 3.2.3).
3. Parameters for modeling turbine-governors in dynamic simulation
programs are usually supplied on a per unit base that is based on the
turbine megawatt rating. If the programs uses the same per unit base for
the generator and turbine-governor model (i.e. generator MVA base)
then care must be exercised to properly represent the turbines maximum
megawatt capability using the turbine-governor output limits (typically the
parameter Pmax) and steady state droop.

3.2.2 Recommendations Related to System Monitoring and for Ensuring


Adequate Governing Response
Providing monitoring capability offers many advantages. These include an
ability to compare simulated and actual unit performance, validating unit
governing response without the need for staged testing of the generator for
verifying the governor performance, and also a means of monitoring on-line the
responsiveness of a unit to primary frequency control. This would make it
practical to provide for a "primary governing frequency control" ancillary service
market. In light of the observed decline in system governing response in parts of
the US, and the advent of the deregulated market, the creation of an ancillary
service market for primary governing seems a prudent approach to hold
participants accountable within such an ancillary market. A methodology for
on-line monitoring to confirm the generating facilities performance in this regard
would be needed. In view of these observations, the following specific
recommendations are made:
1. As shown in several case studies presented in Chapter 2, the ability to
compare and subsequently tune the simulation models against recorded
system response to large disturbances has significantly helped to resolve
some of the disparity between simulation and actual system response.
Therefore it is recommended that tracking the governing performance of
individual generating units on-line be considered. Units above a
designated MW rating (depending on the size of the interconnected
system) should have their gross or net MW output and station frequency
monitored continuously. The unit's MW setpoint should also be monitored.
A minimum sampling rate of one point per tenth of a second is
recommended, with data storage capacity allowing ample time for plant

3-5
personnel to download any data captured during a major disturbance.
The monitored data should be available to appropriate transmission
planning personnel after system frequency disturbances. Thus, a base of
monitoring experience may be accumulated for each unit that will serve
as a basis in determining appropriate adjustments to modeling data that
can then be factored into the simulation databases. This monitoring
capability has many other benefits as well; for example, it can track
compliance with generator under-frequency performance requirements
and will greatly aid post-disturbance analyses.
2. It is recommended that the transmission system authorities should a)
determine the needed amount of primary frequency control to meet
reliability standards, b) determine the monetary value of governing, c)
pay generators for governing, and d) monitor governing performance to
either reward or penalize performance.
3. When using computer simulations, all modeling improvements should be
undertaken uniformly and carefully throughout an interconnected system,
to ensure that results do not deviate further from actual system behavior
with a modeling "improvement" in only one portion of the
interconnection. [5], [52].
4. It is recommended that either testing or monitoring be performed on
generating units above a designated MW rating (depending on the size
of the interconnected system) to determine governor performance and
to determine the mode of operation of the unit (e.g. base-loaded, on
temperature limit, on outer-loop megawatt control etc.), on a routine
basis. How often should be determined by the regional reliability
organization. This information should be reported to the appropriate
transmission system planning staff. One means of achieving this goal
would be through continuous on-line monitoring, as recommended
above under item 1.
3.2.3 Recommended Future Work
Based on the work of this Task Force, and the conclusions reported above, the
following activities for future work are advisable:

1. A Task Force should be formed to identify and make recommendations


for resolving some of the issues identified in this report related to primary
frequency regulation. The risk to security, assumed to be adequate due
to present levels of PGFR by expert judgment, should be confirmed by
technical analysis. The following question should be answered: "If some
portions of an interconnection requires greater PGFR due to their topology
and risk of separation, how does this greater PGFR affect reliability when
the interconnection experiences a disturbance and remains intact? Can
imbalances in PGFR result in increased risk during disturbances?"
2. A Task Force should be formed to review and make recommendations
related to the use of models for turbine-governors for power system
simulations. Although recent documents [6], [1], [16] and other work have

3-6
provided newly developed models for use in modeling thermal turbine-
governors, modern combined-cycle power plants and hydro-turbines,
there may be some benefit to reviewing these recommendations and
consolidating them with older models that still exist in commonly used
simulation programs. The Task Force should make general
recommendations on which models are applicable for which tasks and
which models are presently considered obsolete. Also, the Task Force
should identify any deficiencies in existing models and make
recommendations on improving them.
3. A Task Force should be formed to review and make recommendations on
possible ways of performing on-line monitoring to estimate the governor
steady state droop and response of turbine-generator governor action.
Before new Task Forces are created, the several organizations that are now
involved in these areas, e.g., IEEE, CIGRE, NERC, UCTE, UWG, should be
consulted. Coordination with these groups will ensure that the work scope(s) are
consistent. The scopes of each of the new or revised Task Forces should include
the requirement that these Task Forces continue the coordination, whether
through joint memberships, formal or informal liaisons, or other means as appear
appropriate.

3-7
4 References
1. CIGRE Technical Brochure 238, Modeling of Gas Turbines and Steam Turbines in
Combined-Cycle Power Plants, prepared by CIGRE TF C4.02.25, December 2003.
2. P. Kundur, Power System Stability and Control, McGraw-Hili, Inc., 1994.
3. EPRI Technical Report, NP-2849, "Nuclear Plant Response to Grid Electrical
Disturbances", February 1983.
4. R. P. Schulz, "Modeling of Governing Response in The Eastern Interconnection",
Proceedings of the IEEE PES Winter Meeting, 1999 "Symposium on Frequency
Control Requirements, Trends and Challenges in the New utility Environment"
5. EPIC Engineering, Inc., S. Virmani, "Impacts of Governor Response Changes on
the Security of North American Interconnections", EPRI Report no. TR-1 01080, Oct.,
1992. (RP2473-53)
6. L. Pereira, J. Undrill, D. Kosterev, D. Davies, S. Patterson, "A New Thermal Governor
Modeling Approach in the WECC," IEEE Trans. Power Systems, vol. 18, Issue.2, pp.
819-829, May 2003.
7. NPCC Working Group on Review of Reliability of Inter-Area Operation (COSS-2),
Final Report, May 1994.
8. T. Leung and J. Peet, "Control System Retrofits to Improve Plant Efficiency",
Canadian Electrical Association: Thermal Generating Station Construction and
Commissioning Session, March 1995, Vancouver, Canada.
9. L.M. Hovey, "Optimum Adjustment of Hydro Governors on Manitoba Hydro
System," Trans. AlEE, Vol. PAS-81, pt. 3, pp 581-587, December 1962.
10. F. R. Schleif, A. B. Wilbor, "The Coordination of Hydraulic Turbine-governors for
Power System Operation", IEEE Trans.., Vol. PAS-85, No.7, pp. 750-758, July 1966.
11. P. L. Dandeno, P. Kundur and J. P. Bayne, "Hydraulic Unit Dynamic Performance
Under Normal and Islanding Conditions - Analysis and Validation", IEEE Trans.,
Vol. PAS-97, pp. 2134-2143, November/December 1978.
12. J. C. Agee, H. D. Vu, J. Volk, "Installation and Testing of a Prototype Digital
Governor at Mt. Elbert Powerplant", Proceedings of the IEEE Power Engineering
Society Winter Meeting 1999, pp. 102-107.
13. D. G. Ramey, J. W. Skooglund, "Detailed Hydrogovernor Representation in System
Stability Studies," IEEE Trans. Vol. PAS-89, No.1, pp. 106-112, Jan. 1970.
14. J. M. Undrill and J. L. Woodward, "Nonlinear Hydro Governing Model and
Improved Calculation for Determining Temporary Droop", IEEE Trans., Vol. PAS-86,
pp. 443-453, April 1967.
15. IEEE Working Group on Prime Mover and Energy Supply Models for System
Dynamic Performance Studies, "Hydraulic Turbine and Turbine Control Models for
System Dynamic Studies," in IEEE Trans. Power Systems, Vol. 7, No.1, pp. 167-179,
Feb. 1992.
16. D. Kosterev, "Hydro Turbine-Governor Model Validation in the Pacific Northwest,"
IEEE Trans Power Systems, Vol. 19, No.2, pp. 1144-1149, May 2004.

4- 1
17. M. M. Merrian and D. J. Vandewalle, "Effect of Grid Frequency Decay Transients
on Pressurized Water Reactors", IEEE Trans. PAS, Vol. PAS-95, No.1,
January/February 1976, pp. 269-274.
18. ANSI/IEEE Standard C37.1 06-2003.re, "IEEE Guide for Abnormal Frequency
Protection for Power Generating Plants", IEEE 2003.
19. P. Pourbeik, "The Dependence of Gas Turbine Power Output on System Frequency
and Ambient Conditions", paper 38-101, Proceedings of CIGRE Session 2002,
August 2002, Paris, France.
20. K. Kunitomi, A. Kurita, H. Okamoto, Y. Tada, S. Ihara, P. Pourbeik, W. W. Price, A. B.
Leirbukt and J. J. Sanchez-Gasca, "Modeling Frequency Dependency of Gas
Turbine Output", Proceedings of IEEE PES Winter Meeting, Jan 2001.
21. CIGRE Technical Brochure, Modeling and Dynamic Behavior of Wind Generation
as it Relates to Power System Control and Dynamic Performance, prepared by
CIGRE WG C4.601 on Power System Security Assessment, January 2006.
22. G. Lalor, A. Mullane and M. O'Malley, "Frequency Control and Wind Turbine
Technologies", IEEE Transactions on Power Systems, Vol. 20, No.4, pp. 1905-1913,
November 2005.
23. Cost of Providing Ancillary Services from Power Plants - Volume 1: A Primer, EPRI
TR-1 07270- V1, 4161, Final Report, March 1997.
24. Cost of Providing Ancillary Services from Power Plants - Regulation and Frequency
Response, EPRI TR-107270-V2, 4161, Final Report, April 1997.
25. Frequency Response Standard White Paper, Prepared by the Frequency Task
Force of the NERC Resources Subcommittee, April 6, 2004,
ftp://www.nerc.com/pub/sys/all_updl/standards/sar/Frequency_Response_White
_Paper.pdf.
26. Le-Ren Chang-Chien, Naeb-boon Hoonchareon, Chee-Mun Ong and Robert A.
Kramer, "Estimation of fi For Adaptive Frequency Bias Setting in Load Frequency
Control", IEEE Trans. PWRS, May 2003, pp. 904-911.
27. Naeb-boon Hoonchareon, Chee-Mun Ong and R.A. Kramer, "Feasibility of
Decomposing ACE1 to Identify the Impact of Selected Loads on CPS1 and CPS2",
IEEE Trans. PWRS, Aug. 2002, pp. 752-756.
28. Naeb-boon Hoonchareon, Chee-Mun Ong and R.A. Kramer, "Implementation of
an ACE1 Decomposition Method", IEEE Trans. PWRS, Aug. 2002, pp. 757-761.
29. H. F. lilian, Defining, Measuring and Valuing Frequency Response, Prepared for
Garland Power and Light and the Texas Nodal Team, January 1, 2005.
30. E. H. Allen and M. D. llic, "Reserve Markets for Power Systems Reliability", IEEE
Trans. PWRS, Vol. 15, No.1, February 2000, pp. 228-233.
31. ERCOT web site, www.ercot.com.
32. ERCOT EIA-411 report dated 4-15-2004.
33. EReOT Protocols (www.ercot.com)
34. ERCOT Operating Guides (www.ercot.com)
35. J. W. Feltes and J. R. Willis, "Boiler Effects on Steam Turbine Response", Power
Technologies, Inc., August 1999.

4-2
36. J. Ingleson and M. Nagle, "Decline of Eastern Interconnection Frequency
Response", 1999.
37. S. Niemeyer, "Control Performance", Nineteenth Annual ERCOT Operations
Seminar, April 2002.
38. CIGRE 38.02.14 final report" Analysis and Modeling Needs of Power Systems Under
Major Frequency Disturbances", 1999.
39. Comanche Peak Switchyard Protective Relaying Incident Report, Mark Carpenter,
May 30,2003 (Oncor Electric Delivery Company internal report)
40. M. Henry, "May 15,2003 Event Review", Twenty-First Annual ERCOT Operations
Seminar.
41. S. Lindahl, "Frequency control in the synchronous Nordel system", Nordel annual
report 1983.
42. N. Cohn, "Recollections of the Evolution of Realtime Control Applications to Power
Systems", Automatica, vol. 20, no. 2, pp. 145-162, 1984.
43. R. Kearsley, "Restoration in Sweden and Experience Gained from the Blackout of
1983", IEEE Transactions on Power Systems, vol. PWRS-2, no. 2, pp. 422-428, May
1987.
44. K. Waive, "Modelling Power System Components at Severe Disturbances", Report
38-18, CIGRE Session, Paris, August-September 1986.
45. S. Lindahl, G. Runvik and G. Stranne, "Operational Experience of Load Shedding
and New Requirements on Frequency Relays", Paper presented at the
International Conference on Developments in Power System protection,
Nottingham, 25-27 March, 1997, lEE Conference Publication No. 434, lEE, 1997.
46. S. Larsson and E. Ek, "The black-out in southern Sweden and eastern Denmark,
September 23,2003", Proceedings of the IEEE PES General Meeting, June 2004.
47. L. Pereira, "New Thermal Governor Model Development - Impact on Operation
and Planning Studies on the Western Interconnection" IEEE Power & Energy
magazine, May-June 2005, Vol.3, No.3
48. General Electric Company, PSLF Simulation Program Reference Manual, 2004
http://www.gepower.com/energyconsulting/en_us/pdf/ pslf_manual.pdf
49. L. Pereira, D. Kosterev, D. Davies and S. Patterson, "New Thermal Governor Model
Selection and Validation in the WECC" IEEE Trans. PWRS Feb. 2004, pp. 517 - 523
50. S. Patterson, "Importance of Hydro Generation Response Resulting from the New
Thermal Modeling - and Required Hydro Modeling Improvements", Proceedings
of the IEEE PES General Meeting, Denver July, 2004.
51. WECC White Paper, "Frequency Responsive Reserve Standard", April 2005,
http://www.wecc.biz/documents/librarv/RITF/FRR White Paper v8r clean.pdf.
52. C. Concordia and R. P. Schulz, "Appropriate Component Representation for
Simulation of Power System Dynamics", IEEE Monograph 75 CH0970-4-PWR, pp.
16-19,1975.
53. J. Ingleson and D. Ellis, "Tracking the Eastern Interconnection Frequency
Governing Characteristics", Proceedings of the IEEE PES General Meeting, June
2005, pp. 1461-1466.

4-3
54. L. H. Flink and K. Carlsen, "Operating Under Stress and Strain", IEEE Spectrum, Vol.
15, pp. 48-53, March 1978.

4-4
APPENDIX A: List of Acronyms and Terminology
l/R (Inverse of steady state droop) - represents the speed governing
response of the power system or part of the system under
consideration, caused by control action to increase or decrease
power output of the generating unit(s) in that system in response
to frequency changes. Common units are: MW /Hz. or per unit
power/per unit frequency.
AC Alternating Current
AECB Atomic Energy Control Board of Canada
AEP American Electric Power Company
AGC Automatic Generation Control. A control system, operating
centrally within an all, or a portion of an AC electric power system
that allocates the generation levels of several generating entities,
to meet demand, interchange and frequency commitments.
J3(beta) (J3 = 1/R + D) represents the combined load and governing
response of the power system or part of the system under
consideration. Common units: MW/Hz. MW/mHz. or per unit
power/per unit frequency.
BWR Boiling Water Reactor
CCPP Combined Cycle Power Plant
CIGRE International Conference on Large High Voltage Electric Systems
or Conference des Grands Reseaux Electriques a Haute Tension
CPS1 NERC Control Performance Standard 1
D(damping) Represents the load response of the power system or part of the
system under consideration, caused by the change in power
usage by the loads in that system due to variation of frequency.
Common units: MW /Hz, MW /mHz, or per unit power/per unit
frequency. A common value of D is approximately 1.5% of the
connected load per 1.0% change in frequency, or D = 1.5 per unit.
Droop (R) Steady state droop is a control setting, defined in standards as the
change in frequency required to cause a change in prime mover
power output from full rated output to zero output. The most
common setting for droop is R = 0.05 per unit power per unit
frequency (or 5%), which is 3 Hz for a 100% change in MW output
power on a 60 Hz. system or 2.5 Hz. on a 50 Hz. system.
EI Eastern Interconnection of the United States and Canada.
EMS Energy Management System.
EPRI Electric Power Research Institute
ERCOT Electric Reliability Council of Texas. ERCOT operates as a separate
synchronous interconnection of several utilities in Texas. ERCOT is
a part of NERC.
HRSG Heat Recovery Steam Generator
HVDC High-Voltage Direct-Current
IEEE The Institute of Electrical and Electronics Engineers
ISO Independent System Operator
MAPP Mid-Continent Area Power Pool

A-I
NERC North American Electric Reliability Corporation (www.nerc.com).
This is the North American electric reliability organization. NERC
covers four large synchronous systems that are connected by
HVDC systems: Eastern US and Canada, Western Canada and US
(WECC), Texas (ERCOT), and Quebec.
NORDEL Nordic Electricity. NORDEL is a body for cooperation between the
transmission system operators (TSOs) in the Nordic countries of
Denmark, Finland, Iceland, Norway and Sweden.
Normal System The most common operating mode within a balancing authority
Operating or control area of an interconnection, or of a portion of an
Condition interconnection. In this mode, all connected load is being served,
all negotiated and scheduled power interchanges are being met,
all transmission system parameters (e.g., frequency, voltage, flows)
are within acceptable bounds [54].
NPCC Northeast Power Coordinating Council. The bulk power
coordinating body for the New England, New York, Ontario, and
Quebec Areas of the - Eastern Interconnection of the United
States and Canada. NPCC is a part of NERC.
PGFR Primary Frequency Regulation - system frequency to turbine
governing controls and load sensitivity to frequency
PMU Phasor Measurement Unit
PWR Pressurized Water Reactor
Secondary Is dispatching to shift generation among available units to meet
frequency system demand and to achieve several purposes, which may
regulation including balancing generation to electric energy supply
commitments (often called interchange or tie-line control),
maintain frequency at the scheduled value (50 or 60 Hz) and/or
minimizing the cost (often called economic dispatch). Secondary
frequency regulation may be done manually or automatically,
called Automatic Generation Control (AGC). By intent,
secondary frequency regulation operates more slowly than speed
governing, to prevent interactions between these controls.
TSO Transmission System Operator
Turbine Governing The automatic response of a generating unit (or of units in the
aggregate) to increase (or decrease) power generation in
response to decreasing (or increasing) frequency.
UCTE The Union for the Coordination of Transmission of Electricity. This is
the association of transmission system operators in continental
Europe.
UFLS Under-Frequency Load-Shedding
UWG Utility Working Group
WECC Western Electricity Coordinating Council. WECC is a separate
synchronous interconnection of utilities in the US, Mexico and
Canada, generally west of the Rocky Mountains.
PGFR Primary frequency regulation - system frequency to turbine
governing controls and load sensitivity to frequency

A-2

Anda mungkin juga menyukai