net/publication/34466506
CITATIONS READS
29 504
1 author:
Mahesh S. Illindala
The Ohio State University
96 PUBLICATIONS 1,111 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Mahesh S. Illindala on 08 July 2014.
by
Mahesh S. Illindala
Doctor of Philosophy
(Electrical Engineering)
at the
UNIVERSITY OF WISCONSINMADISON
2005
© Copyright by Mahesh S. Illindala 2005
All Rights Reserved
i
Abstract
Distributed Generation (DG) systems are being increasingly favored for meeting
customers. Many DG systems utilize a voltage source inverter (VSI) based on pulse
width modulation (PWM) as the utility/load interface and therefore are able to offer
benefits like power quality conditioning in addition to providing of utility grade power.
This thesis addresses the control aspects of these inverter based DGs enabling them to
meet the wideranging load demands of the highend customers by their integration with
The control of VSI is undertaken in two stages – viz., a multiloop internal control
using stationary frame vector regulators and an external generation control. The vector
regulators are designed on threephase complex spacevector quantities. They are used to
regulate the load terminal voltage under balanced as well as unbalanced conditions. High
quality sequence filters are incorporated into the controller for accurate compensation of
transformer impedance effects. The generation control is carried out by means of active
operation of the microgrid with load sharing by the DGs and seamless interchange
are made on chain and parallel topologies. Guidelines are provided for design of the
active powerfrequency and reactive powervoltage controllers to meet the IEEE P1547
Acknowledgements
I would like to express my sincere gratitude and thanks to advisor Professor Giri
Venkataramanan for his guidance, counsel and encouragement at every stage of this
research. I thank the National Renewable Energy Laboratory (NREL) and Wisconsin
Electric Machines and Power Electronics Consortium (WEMPEC) for funding this
I have been fortunate to work with the WEMPEC professors, students and support
staff who provided an excellent working environment with all the resources that made
I acknowledge the cooperation and support of Paolo, Howard, Mike, Nick, Sean,
Paul, Raahul and Shashank who helped me to their fullest with the laboratory microgrid.
While working in the WEMPEC laboratory, Raymond Marion always had helpful
/labmates and a few others  Jackson, Panda, Paolo, Rajesh, Amit, Yongsug, Jun, Bin,
Yusuke, Metin, Rick, Sreesha, Shashank, Sandeep, Bhageerath, Bala, Brian and Andrew.
patience and assistance during the significant part of this research work. In addition, I
Table of Contents
Abstract ........................................................................................... i
1.1 ......General.........................................................................................................1
Control Systems.....................................................................................................32
2.7 ......Summary....................................................................................................49
Regulator 52
3.1 ......Introduction................................................................................................52
3.2 ......Plant Transfer Function Modeling of the DR Filter Interface Network ....53
Functions................................................................................................................73
3.6 ......Summary....................................................................................................86
Microgrid 89
4.1 ......Introduction................................................................................................90
4.6 ......Summary..................................................................................................127
Microgrid .................................................................................................146
Microgrid .............................................................................................................152
5.4 ......Summary..................................................................................................156
(DR) 189
7.3 ......Summary..................................................................................................192
Bibliography................................................................................193
206
List of Figures
Figure 23 Bode plot of loop gain TL_1(s) for (a) negative and (b) positive
sequence components of the space vector........................................29
Figure 25 Bode plot of the loop gain TL1(s) (solid line) and closedloop
gain GCL1(s) (dashed line). (Only positive sequence response
is shown.) .........................................................................................33
Figure 27 Bode plot of the loop gain TL2(s) (solid line) and closedloop
gain GCL2(s) (dashed line). (Only positive sequence response
is shown.) .........................................................................................36
Figure 210 Variation of peak overshoot and settling time for vector
controlled secondorder ac plants.....................................................38
Figure 34 Block diagram of the effective plant transfer function of the
DR filterinterface network under unbalanced/interconnected
DR conditions for space vector quantities .......................................55
Figure 36 Block diagram of the current regulator for (a) space vector
quantities, and (b) orthogonal αβ coordinate quantities ..................59
Figure 37 Bode plot of the loop gain for the current regulator loop ...................60
Figure 39 Bode plot of the loop gain for the voltage regulator loop...................63
Figure 310 Bode (gain) plot of the output impedance (a) without the
multiloop controller and (b) with the multiloop vector
controller. .........................................................................................65
xii
Figure 312 Bode plot of the transfer function GNSF(s) for (a) negative
sequence and (b) positive sequence components of the space
vector................................................................................................68
Figure 314 Bode plot of the loop gain for the negative sequence filter
(NSF) Gfn(s) for (a) negative sequence and (b) positive
sequence components of the space vector iLt(t) ...............................70
Figure 316 Bode plot of the negative sequence filter GNSF(s) for different
values of gain ωo (20, 200 and 2000). (a) negative sequence
and (b) positive sequence components of the space vector..............73
Figure 318 Nyquist path on the splane and the TL locus for M =
ωo/(2ω60)...........................................................................................75
Figure 319 Nyquist path on the splane and the TL locus for M =
jωo/(2ω60) .........................................................................................76
Figure 320 Bode plot of loop gain TL(s) for M = ωo/(2ω60) for (a)
negative and (b) positive sequence components of the space
vector................................................................................................78
Figure 321 Bode plot of loop gain TL(s) for M = jωo/(2ω60) for (a)
negative and (b) positive sequence components of the space
vector................................................................................................78
Figure 323 Operation of the voltage control loop for vCf with the
response vCf for a complex exponential step input voltage
reference vCf†....................................................................................81
xiii
Figure 325 Load terminal lineline voltage ac & bc (top two) and load
current in a & b phases (bottom two) for a balanced 10 kW
resistive load.....................................................................................82
Figure 41 Voltage regulated DR’s analogy to a voltage source for power
studies...............................................................................................91
Figure 42 Generation of voltage command for the inverter based DR ...............92
Figure 47 Block diagram of the proposed reactive power controller for a
DR ....................................................................................................99
Figure 55 Locus of the dominant eigenvalue of the real power flow in a
5DR parallel microgrid as ωG5 is reduced below 1 rad./s. ...........148
Figure 61 One line diagram of the laboratory microgrid consisting of two
DRs.................................................................................................160
Figure 62 Circuit schematic of the PWM inverter based DR unit with its
LCL filter .......................................................................................163
Figure 63 Wiring layout of the transformer, loads and the tieline
connections to the busbar (to be modified for contactor)...............166
Figure 64 Distributed generator DG1 supplying power to its local load as
well as tied to the microgrid...........................................................168
Figure 65 Connections of the distributed generators DG2 and DG3 to the
microgrid ........................................................................................169
Figure 68 Operation of the voltage control loop for vCf with the response
vCf for a complex exponential step input voltage reference
vCf†. DSP obtained from PC Master™ of Metrowerks®
CodeWarrior™. ..............................................................................173
Figure 610 Load terminal lineline voltage ac & bc (top two) and load
current in a & b phases (bottom two) for a balanced 10 kW
resistive load. Current is scaled at 1 A = 1 V.................................174
Figure 614 Transformer primary current, PSF output and NSF output
waveforms in the PC Master™ for an unbalanced load.................178
Figure A3 Cell discharge voltage Vs. time for a leadacid battery [101]..........212
Figure A4 Peak current and voltage per cell of a leadacid battery for a
highrate discharge [101] ...............................................................213
List of Tables
Chapter 1 Introduction
1.1 General
nuclear plants with large generation capacities are located at distant locations from load
centers. The energy is transported from the generation plant to the loads through
equipment are becoming utilized to their maximum capacity. In the meanwhile, the
electrical power sector is continuing to witness a steady growth in the load demand. This
increased load demand requires not just an increase in generation capacity, but also an
transmission and distribution network is already massive and widespread, the constraints
of space and right of way have made the system planners to look for alternative avenues
for meeting the excess electrical energy requirements for these loads. One approach that
is favored is the use of distributed generation [1]. Distributed Generation using a variety
that are distributed in a power network, to supply the electric power needs of electrical
dispatchability [2]. The capacity of these generators may range between a few kW
through a few MW. These generators are located on the utility system, at the site of the
utility customer, or at an isolated site not connected to the power grid. The DG is said to
operate in the gridinterfaced mode when it is connected to the utility, and is said to
turbine, combined cycle turbine, lowhead hydro, fuel cells and renewable sources such
as wind and solar power. While some of these technologies make use of rotating
machines, others employ a power electronic converter to derive utility grade ac power
with a primary energy source, dc link and a dcac converter (or power electronic inverter)
rotating masses, they do not have the inertia that would source or absorb transients in
load demand. For this reason, the inverter’s dc link includes an energy storage device to
meet load demands while the primary source availability and/or dynamics precludes load
following [10]. The energy storage feeds to the load requirements during startup or
transient in load demand, until the primary energy source can meet the load demand. The
3
inverter based DGs when employed with suitable controls, are capable of providing a
wellregulated voltage supply against imbalances and ridethrough against short rated
power quality events at the sensitive load bus. When coordinated with proper protection
loads. Thus, use of power electronic converter enables DG systems to provide premium
power [9,10], by locating them near the critical loads. In view of this, they are sometimes
3 phase
micro dc dcac
source link converter
ac output
Figure 11 Block diagram of a typical inverterembedded distributed resource (DR) unit
Uninterruptible Power Supply (UPS) system, the extent to which they will be used to
provide generalpurpose utility grade ac power requires their control and protection
active filtering, voltage sag mitigation etc [13]. When such ancillary functions are added
to the regular generation control in its operation, the DR becomes a more viable choice as
it provides a reliable and high quality power supply at a reasonable cost [14].
The typical inverter embedded DG system shown in the form of a block diagram
in Figure 11, is illustrated by a threephase circuit schematic in Figure 12. Each such
4
unit is rated for supplying threephase loads or singlephase loads at the power frequency,
typically 60 Hz. As seen in the figure, the dc bus is modeled as a stiff dc source (Vdc) in
the presence of adequate dc energy storage to manage load tracking [15] (refer Appendix
A on battery energy storage requirements for DR systems). The Insulated Gate Bipolar
Transistors (IGBTs) and diodes illustrated in the figure are typically operated under
utilized to essentially provide a neutral terminal for the single and threephase loads. Use
of the transformer also (a) assists in steppingdown the voltage to a level of the load
rating, (b) it isolates the zero sequence currents on the load side of the network, thus
protecting the inverter during lineground faults and furthermore, (c) disallows dc
injection from the DR into the network, and (d) provides additional filtering through its
leakage impedance. The generic impedances (Za, Zb and Zc) model the load, representing
the load bus at the transformer secondary may be interfaced to utility grid and/or another
DR bus, in which case the current flowing out to the rest of the system is labeled as Itie.
5
VNf Itiea
I tieb VanL
Itiec VbnL
VcnL
ILda I Ldb I Ldc
TAC Ta TBA Tb TCB Tc
Za
Zc
Zb
VnL
Figure 12 Circuit schematic of the threephase PWM inverter based DR connected to a load
phase and threephase loads as well as those that are sensitive to disturbances in power
quality, in a typical distribution system. While the passive loads, viz., resistive, inductive,
electronics are largely nonlinear in nature as they draw nonsinusoidal currents from a
sinusoidal voltage supply. Small capacity loads are typically singlephase and the large
capacity ones are predominantly threephase loads. Sensitive loads employ hitech
6
machinery like computers and electronic control equipment that are susceptible to power
quality disturbances.
Among linear loads, those operating at low power factors are known to cause a
great burden to the utility for providing reactive support. Harmonic currents drawn by
nonlinear loads can produce losses and overheating of electric distribution system
terminal voltage. Recently, it has also been reported that the common power quality
problems such as voltage sags, voltage swells, voltage spikes and shortterm outages are
estimated to cost the US economy $26 billion annually [16]. The losses are due to loss of
productivity in the downtime of sensitive loads caused by the power quality events. Thus,
it would be highly desirable for DG systems to mitigate power quality disturbances and
The traditional approach to obtaining high quality power supply for sensitive
loads involves use of UPS systems for conditioning the input voltage. However, the
problems in battery management at high power levels, high capital costs and minimal
utilization under normal operating conditions are some of its drawbacks. These systems
also contain transfer switches whose operation is critical to providing continuous power
to loads [14]. Backup generation systems based on rotating machines can be used as an
alternative system to supply power for sensitive loads during temporary interruptions in
power supply. However, they have a slower response time and also require automatic
transfer switches for a change over between power sources in case of power quality
7
events. During the transfer between main supply and backup, there may be a loss of
continuity of power to the loads and hence these systems are generally unable to meet
stringent power quality indices. Moreover, they cause air pollution and may not be used
for more than a limited number of hours per year due to limits on exhaust emissions [17].
While the inverter embedded DG systems can be made free from such concerns,
the power rating of the primary energy source they employ limits their size. This
necessitates clustering of multiple small rated DRs for serving under any load condition.
Moreover, use of multiple smallrated DRs helps in reducing the investment in meeting
when interconnected and dispersed such that they are located near critical loads can
improve the reliability of the power supplied to the critical/sensitive loads. Such a DG
in Figure 13. Under normal operating conditions, the DRs, viz., DR1 and DR2, share the
total load in the system along with the utility grid supply. However, when the utility
supply is down, the DRs are rated to provide all the energy requirements of atleast the
critical loads.
As seen in Figure 13, the interconnection between any two DRs is made by
synchronizing and control logic circuitry. Likewise, a threephase static switch consisting
of thyristors is utilized for the interconnection between the microgrid and the grid mains.
Before connection, the threephase voltages on both sides of the switch (contactor or
static switch) are matched by the synchronizing logic circuitry. Not shown in the Figure
commissioning, safety, and maintenance at the point of common coupling (PCC) of the
DR interconnection to the Electric Power System (EPS) are laid out in the IEEE Standard
Load1
T1 SS 1 T2 Lt1 Lf1
Δ Y YΔ Cf1
DR1
B1
Grid
Interconnect
T ieline
Interconnect CN1
Load12
B2
B Busbar Load2
C Capacitor
CN Contactor w/ Manual Override T3
L Inductor Lt2 Lf2
R Resistor
SS Static Switch w/ Manual Override
YΔ Cf2
T T ransformer DR2
B3
mode. The individual inverters should be capable of sharing active and reactive power in
a predetermined manner and circulating power should be avoided. In order to realize this,
it is essential to have good control of the power angle and the voltage level by means of
the inverter. Control of the inverter’s frequency dynamically controls the power angle,
and the flow of the real power. On the other hand, the flow of reactive power can be
ensure that the inverter takes up load changes in a predetermined manner, without
but must not be critical for system operation. Essentially this implies that the inverter
Over 90% of voltage disturbances in the utility lines are singlephase voltage sags
caused by momentary line to ground faults in distribution systems. Hence the control
strategy of the inverter should be able to meet with situations when the utility grid having
residual imbalances in voltage and when the load system has imbalances in current.
may consist of nonlinear loads such as rectifiers. The control strategy of the inverters
should be such that they are capable of supplying real and reactive power demands as
10
well as the harmonic currents necessary to feed such loads without exacerbating the
situation.
(i) inverter internal controls to regulate the load bus voltage even amidst
Firstly, the control of VSI employed in the DR is to regulate the load bus voltage
even in the event of imbalances and enable interconnection to other generation sources. A
stationary frame vector regulator has been designed to regulate load terminal voltage
under unbalanced conditions. This regulator is equivalent to the complex vector regulator
opposite directions. Secondly, research presented here extends the work to address the
mechanisms for active and reactive power sharing based on frequency droop control and
voltage droop control by providing design guidelines to meet IEEE P1547 performance
programmable operation in two operation modes, viz., normal grid connect mode and
intentional island mode [22,23]. In the normal gridinterfaced mode of operation, the DR
inverter is connected to the loads in parallel with the grid, and it is controlled as a current
source with the grid providing the voltage as well as frequency references. During grid
supply interruptions, the DR controller senses the loss and immediately disconnects from
both the grid and the loads. When the utility grid returns to within its specified limits, the
DR may be programmed to restart and recommence power to the connected loads. On the
other hand, when the DR is operated in the intentional island mode, it is programmed as a
voltage source based on openloop voltage control with fixed voltage and frequency
reference. This control strategy causes interruption of power supply to loads as the DR
power needs to be recycled for switching between the two modes of operation.
In inverter based wind energy systems, the control strategy generally adopted
involves current control to output the desired real power in the normal gridinterfaced
mode of operation [24]. These systems require additional information of the grid voltage
through a phaselocked loop (PLL) to generate the reference currents for a desired power
output. It is generally assumed that the grid voltage is stiff so that the real and reactive
overcome the limitations in the existing technology. It regulates the load terminal voltage
to within the specified limits and allows a seamless operation in both gridinterfaced and
standalone modes. Furthermore, power quality conditioning capability is added into the
12
control strategy to take complete advantage of the power electronic inverter used as the
utility/load interface. The literature reported relevant to the work presented in this thesis
motor drives and harmonic/sequence filtering has given a tremendous boost to the design
of controllers for voltage source inverter (VSI) based distributed resources (DRs). The
performance is the common link among all these applications. For that reason, the
literature presented here includes various advances made in UPS, ac motor drives,
uninterruptible power supplies (UPS) or motor drive systems, by and large feeding
balanced loads [21,2533]. In the stationary frame UPS control, the stateofart
provide poor performance except at high bandwidth as they do not give a zero steady
state error for ac sinusoidal quantities. For this reason, the design philosophy is generally
based on regulators operating in the synchronously rotating dq reference frame that have
been long established for controlling VSIs [25,26,21]. They require transformation of ac
quantities into a synchronously rotating dq reference frame to form dc quantities and then
13
ambient industrial or residential load conditions that have a broader degree of variability
[34]. These conditions include persistent imbalance in loads, especially due to single
phase distribution circuits, and steadystate and transient imbalances in the utility grid
voltage [34,35]. Under such conditions the system requires implementation of separate
controllers in positive and negative rotating reference frames that collectively supply the
reference voltage of the inverter, thereby complicating system design further [37].
Recently, some authors have reported regulators in the stationary reference frame that
would bring about zero steadystate error and give a response similar to the synchronous
regulators for single and threephase UPS that decouple the LC filter dynamics of the
analytical and computational advantages. It has been applied in the phasor dynamic
regulators that are devised on threephase spacevector quantities and realized in the
synchronous frame have been reported [21,36,37,43]. These vector regulators utilize
synchronously rotating reference frame for the regulator design. They implement a more
desirable pole/zero cancellation as they decouple the crosscoupling between the d and q
On the other hand, the H∞ repetitive controller was employed in [46,47,48,49] for
detailed system design [49] and their response time is approximately 5 cycles following a
disturbance [46]. For current regulation in nonlinear loads, Yuan et al [38] and
Allmeling [39] have identified the problems associated with transforming the quantities
into multiple rotating frames and proposed stationary frame transfer functions that enable
because typical operating conditions include persistent imbalance in loads, especially due
to singlephase distribution circuits, harmonics due to likely nonlinear loads and steady
state and transient imbalances in the utility grid voltage. Harmonic filtering concepts
have been originally proposed for active filters. The traditional approach to eliminate the
notch filters at the fundamental frequency. However, they provide an output that contains
severe distorted harmonics. The instantaneous reactive power theory proposed by Akagi
15
et al [50] circumvented this problem, but it however fails to extract current harmonics
linear current waveform to a synchronously rotating frame such that the fundamental
frequency component becomes a dc quantity in this new reference frame. The dc quantity
is filtered by a lowpass filter and the filtered output is transformed back into the
stationary frame. For extraction of individual harmonics, the same approach can be
velocity corresponding to the harmonic and filtering the dc quantity [52]. In a lightly
negative sequence would also output a significant amount of the positive sequence
content. Reduction of the bandwidth, while improving the steadystate filtering, has poor
dynamic response and hence is unacceptable. Cascaded sections of such complex filters
have also been employed and their complex transfer function representation presented
[53]. Similar methods have been applied in electric machine diagnostics as well [54].
However, use of cascaded filters gives rise to phase errors especially when applied at low
Fast stationary frame current regulators for active filters have been proposed by
Yuan et al [38] and Allmeling [39]. They employed generalized stationary frame
integrators to realize equivalent integral controllers for harmonic filtering that would
Sequence Filtering has also been employed in detecting spatial saliencies, i.e.,
16
inductance of the ac machine can be detected from the positive and negative sequence
carrier signal current components drawn by the machine when a balanced high frequency
carrier voltage is applied [55,56]. The threephase currents are transformed to rotating
reference frame and lowpass/highpass filters are employed for filtering of the required
component [57,53,58].
Control
The generation control in VSI based DRs follows the concepts originally
of multiple UPS units to achieve a higher power rating as well as better reliability
through redundancy has been presented by Holtz et al [59]. In this multipleinverter UPS
system, each UPS module has its own controller that communicates with the controllers
for distributed UPS units with no communication between them. The inverters’
ac synchronous machines [61,62], but with no equivalent inertia. Based on the power
ratings of the individual UPS units, design guidelines are formulated [60].
system, having arbitrarily fixed voltage references for the DR can cause excessive
17
reactive power circulation between the generators [63]. This fact has been recognized
to minimize the reactive power flow between the two generators due to discrepancies in
the voltage setpoints, it is necessary to have soft setpoints that can be realized by a
reactive power sharing, a small ac voltage signal is injected into the system as a control
signal. The reactive power to be shared among the paralleled inverters determines the
frequency of this control signal. The small active power due to the control signal is made
use of in adjusting the voltage magnitude of the inverter. Similar approach is used for
generation system. Inverters connected to the same node of the distribution system are
controlled in a masterslave framework with the master having a voltage controller and
the slaves consisting of current controllers with communication between all of them. H∞
repetitive controllers are employed for the voltage/current control. On the other hand,
power sharing for inverters at remote nodes, where it is difficult to have communication,
Piagi [66] presented microgrid controls that enable inverters to track either active
vary the power setpoint is proposed when an inverter based DR’s power limits are
18
reached.
reported that microgrids that were stable with only couple of combustion turbines became
regulator that provides zero steadystate error for threephase systems even under
vector regulators for voltage regulation at the sensitive load bus of a distributed
oriented approach is presented for designing the active powerfrequency droop and the
standalone as well as interfaced modes. Design variables are correlated to the IEEE
laboratory microgrid testbed consisting of two VSIbased DRs are presented for
Chapter 2 describes the characteristics of the complex vector quantities and their
19
Stationary frame based vector controllers are proposed that give zero steadystate error
performance of the vector regulated system are discussed. Realization and comparative
presented.
proposed, based on the stationary frame vector controller, for regulation of the voltage at
sensitive load bus under load as well as line imbalances and faulted conditions. It consists
of an inner current loop for the filter inductor current and an outer voltage loop for the
filter capacitor voltage. State and disturbance feedback decoupling is employed to make
load conditions. Frequency domain analysis techniques based on Bode plots for both
positive and negative sequence quantities are employed for the design of the regulators in
the stationary frame. A high fidelity complex filter is proposed for extraction of
analysis of the complex filter is presented by means of Nyquist and Bode diagrams.
simulated to verify the steadystate and dynamic response of the vector regulator.
for microgrids having a chain topology are also covered in this chapter. Sufficient
stability of n DR microgrids. These studies consider both real and reactive power flows
covered. Guidelines are provided for design of the active powerfrequency droop and
analysis, the IEEE P1547 performance specifications [18] are also utilized in the design
process. Finally, selected results from simulation of DRs equipped with generation
analyzed for the simplest case of a 3DR mesh network based on its analogy to the 3DR
parallel network. Since a nonzero R/X ratio is typical of wires in medium voltage
distribution systems, effects of R/X ratio are investigated in the active and reactive power
flow.
contains the design and specifications of various equipment used. Selected experimental
results demonstrating the performance of the DRs controlled with the internal vector
regulator and external generation controls are also presented in this chapter.
21
Finally, the conclusions of this dissertation and further research are presented in
Chapter 7.
22
complex spacevector quantities. This chapter reviews the complex signals and systems
for threephase power systems and subsequently develops feedback controllers that offer
zero steadystate error at any arbitrary frequency under balanced and unbalanced
characteristics such as peak overshoot or settling time and the phase margin information
Quantities
[44,45]
2
f(t) = 3 {fa(t) + γ fb(t) + γ2 fc(t)} (21)
⎡ fα(t) ⎤ ⎡ fa(t) ⎤
where ⎢ ⎥ = [C] ⎢ fb(t) ⎥ (23)
⎣ fβ(t) ⎦
⎣ fc(t) ⎦
1 1
⎡1 ⎤
and
2⎢
[C] = 3
2 2
⎥ (24)
⎢0 3  3⎥
⎣ 2 2 ⎦
complex spacevector f(t). The trajectory followed by the spacevector f(t) is an ellipse
with the major and minor axes of length (Ap + An) and (Ap – An), respectively. The
fb(t) β Ap  An
fa(t)
t f
φ/2
α
fc(t) Ap + An
Magnitude Phase
Ap
An
ω ω
−ω60 ω60 −ω60 ω60
φ
vector. Likewise, threephase physical systems like ac electrical machines that are multi
transfer functions have also been reported in the fields of communications and signal
processing [69,70].
Consider the transfer function of a linear timeinvariant (LTI) system for space
N(s)
G(s) = D(s) (29)
systems that can be described by an ordinary differential equation (ODE) with constant
coefficients assuming zero initial conditions [71]. For an ODE with constant coefficients
in the real plane R, the transfer function G(s) has poles and zeros that are either real or
phase illustrates this case. The ODE in spacevector variables illustrating the response of
series RL load is
d
v(t) = R i(t) + L dt i(t) (210)
Instead, consider the ODEs for a threephase induction machine in the stator
⎛ d⎞ Md
vs(t) = ⎜Rs + σ Ls dt⎟ is(t) + L dt ψr(t) (211)
⎝ ⎠ r
⎛ d⎞
0 = Rr ir(t) + ⎜ j ωr + dt⎟ ψr(t) (212)
⎝ ⎠
26
where ωr is the rotor speed in electrical rad./s, ψr(t) is the rotor fluxlinkage vector and σ
= 1 – M2/(LsLr).
The ODEs in (211) and (212) have constant coefficients in the complex plane C
if the rotor speed is maintained constant. Accordingly, they represent an LTI system that
complex conjugate pairs. Typically, transfer functions with real coefficients are
formulated for either the real or imaginary parts of the system, rather than viewing the
The next section deals with the development of complex vector regulators
Systems
Briz et al [21] have introduced complex vector regulators for control of balanced
rotating reference frame by Park’s transformation [68]. This section presents an approach
to realize the same complex vector regulator in the stationary reference frame quantities.
complex spacevector quantity variables is illustrated in Figure 22. Assuming the plant
with input fin(t) and output fout(t) to constitute an LTI system, it can be represented by a
transfer function Gp(s) (say). A complex spacevector regulator transfer function Gc(s)
can be designed for specifications such as steadystate error and bandwidth in regulating
the output fout(t) to a desirable value fref(t). Moreover, the controller must provide a high
fin(t)
fdx(t)
fref(t) fout(t)
Gc(s) Gp(s)
TL(s)
Figure 22 Generic block diagram of a feedback controller for an LTI system
Assuming a unity gain feedback, the loop gain of the feedback control system in
Fdx(s) and _{fout(t)} = Fout(s), the Laplace Transform of the output is [71]
TL(s) Gp(s)
Fout(s) = 1 + T (s) Fref(s) + 1 + T (s) Fdx(s) (215)
L L
28
vector regulator transfer function Gc(s) can be determined to achieve a loop gain of
ωo
TL_1(s) = (216)
s – jω60
⎪
In this case, TL_1(s) ⎪⎪ → ∞ (217)
⎪s = +jω60
The closedloop transfer function provides a unity gain at the frequency of ω60
1
GCL_1(s) =
⎛s  j ω60⎞ (218)
1+⎜ ⎟
⎝ ωo ⎠
Figure 23 illustrates the Bode plots of the transfer function TL_1(s) for space
vector quantities with solid lines denoting positive sequence response and dashed lines
denoting negative sequence response. As seen in this figure, the Bode plot displays a
varied frequency response to the positive sequence and negative sequence quantities. The
frequency response for the positive sequence complex exponential quantities is indicated
by the characteristics for ω > 0, and those for negative sequence complex exponential
60 60
40 40
Magnitude (dB)
Magnitude (dB)
20 20
0 0
20 20
40 40
60 60
4 3 3 4
1 .10 1 .10 100 10 1 0.1 0.01 0.01 0.1 1 10 100 1 .10 1 .10
Frequency (Hz) Frequency (Hz)
135 135
90 90
Phase Angle (deg)
0 0
45 45
90 90
135 135
4 3 3 4
1 .10 1 .10 100 10 1 0.1 0.01 0.01 0.1 1 10 100 1 .10 1 .10
Frequency (Hz) Frequency (Hz)
(a) (b)
Figure 23 Bode plot of loop gain TL_1(s) for (a) negative and (b) positive sequence components of the
space vector.
plant transfer function between the applied voltage and the current drawn by the series
RL load is
1
Gp(s) = R + s L (219)
The complex spacevector regulator transfer function for the plant Gp(s) to
TL_1(s)
Gc_1(s) = ^ (220)
Gp(s)
^ ^
ωo(R + s L)
which gives Gc_1(s) = (221)
s – jω60
30
⎛ ^
ωo R⎞⎟ ⎛ 1
⎜ ^ ⎞
Gc_1(s) = ωo L + s ⎜
or ⎝ ⎠ ⎛– jω60⎞⎟ (222)
⎜1 + ⎜ ⎟⎟
⎝ ⎝ s ⎠⎠
⎛ Ki⎞
i.e., Gc_1(s) = ⎜Kp + s ⎟ Gosc_p(s) (223)
⎝ ⎠
ω R
^
^
where Kp = ωo L, Ki = o (224)
s
⎛ 1 ⎞
Gosc_p(s) = ⎜
and ⎛– jω60⎞⎟ (225)
⎜1 + ⎜ ⎟⎟
⎝ ⎝ s ⎠⎠
In the above equations, the estimated parameters of the series RL load used in the
with PI regulator to become capable of tracking positive sequence quantities along with
illustrated in Figure 24. It consists of two integrators along the coupled feedback paths
fα(t) wα(t)
uα(t) ω60 ⌠
⌡
uβ(t) ω60 ⌠
⌡
fβ(t) wβ(t)
Figure 24 A backward coupling form of realization of oscillator Gosc_p(s) in the αβ reference frame
amplitude for only positive sequence quantities in the input even with zero initial
functions for negative sequence quantities, a simple reversal of α and β terminals can
achieve it. As a result, the negative sequence oscillator would have the transfer function
⎛ 1 ⎞
Gosc_n(s) = ⎜
⎛jω60⎞⎟ (226)
⎜1 + ⎜ ⎟⎟
⎝ ⎝ s ⎠⎠
An observation from the Bode plots of loop gain TL_1(s) (refer Figure 23) is the
occurrence of two gain crossover frequencies – one of positive slope and the other of
negative slope. The next section analyses the impact of such occurrences on the dynamic
firstorder plant. The vector regulated system tracks an ac complex exponential input of
frequency ω60 = 2π(60) rad./s. The loop gain of such a unity feedback gain control
system is
Gosc_p(s)
TL1(s) = (227)
⎛s⎞
⎜ω ⎟
⎝ o⎠
⎛ 1 ⎞
Gosc_p(s) = ⎜
where ⎛– jω60⎞⎟ (225)
⎜1 + ⎜ ⎟⎟
⎝ ⎝ s ⎠⎠
Closedloop transfer function of the feedback system with a loop gain TL1(s) is
given by
1
GCL1(s) =
⎛  j ω60⎞
s (228)
1+⎜ ⎟
⎝ ωo ⎠
Bode plots of the loop gain TL1(s) and closedloop gain GCL1(s) for positive
sequence quantities alone are plotted in Figure 25 for ωo = 2π(40) rad./s. In this figure,
the loop gain plot is displayed by a solid line and closedloop gain by a dashed line. As
seen in the figure, the loop gain plot has an infinite magnitude at frequency ω60 = 2π(60)
rad./s and therefore the closedloop gain is unity with zero phase delay at this frequency.
33
Furthermore, the gaincrossover frequencies are displaced equally on either side of the
center frequency ω60 by ±ωo and the phase margin is 90o at both these gaincrossover
frequencies. The loop gain crosses with rising edge of single slope at the lower gain
crossover frequency of (ω60  ωo) with a phase advance of 90o and it crosses with falling
edge of single slope at the higher gaincrossover frequency of (ω60 + ωo) with a phase
delay of 90o.
20 180
135
90
45
0 0
45
90
135 φm
20 180 3
10 100 1 .10
3 10 100 1 .10
Frequency (Hz) Frequency (Hz)
Figure 25 Bode plot of the loop gain TL1(s) (solid line) and closedloop gain GCL1(s) (dashed line). (Only
positive sequence response is shown.)
To determine the complex exponential step response for positive sequence, the
⎧0 for t ≤ 0
cu(t) = ⎨ j ω t (229)
⎩e 60 for t > 0
1
CU(s) = _ {ej ω60 t u(t)} = (230)
s – j ω60
The complex exponential step response f1(t) of closedloop system with transfer
Figure 26 illustrates the complex exponential step response f1(t) for closedloop
transfer function GCL1(s). As seen in the figure, the response is of first order with a
bandwidth of ωo rad./s and a settling time of Ts = 5/ωo seconds. Furthermore, f1(t) has no
1.5
0.5
β
f1α(t) t
1.5
0 0.012 0.024 0.036 0.048 0.06
1 0.5
f1
1
0.5
1.5
α
1.5 1 0.5 0 0.5 1 1.5 1.5
0.5
1
1 0.5
t
1.5 f1β(t)
0 0.012 0.024 0.036 0.048 0.06
0.5
1.5
Figure 26 Complex exponential step response f1(t) of the closedloop transfer function GCL1(s)
While the previous section demonstrated the response of a first order plant, higher
order physical systems produce overshoot and oscillations in their time response. As
35
secondorder plants are the simplest systems that exhibit such a behavior, consider a
closedloop system employing the vector regulator for control of a secondorder plant.
The vector regulated system tracks an ac complex exponential input of frequency ω60 =
Gosc_p(s)
TL2(s) = and (233)
s ⎛ s⎞
⎜
ωo ⎝
1+ ⎟
ω2⎠
1
GCL2(s) = (234)
⎛s – jω60⎞ s ⎛s – jω60⎞
1+⎜ ⎟+ ⎜ ⎟
⎝ ωo ⎠ ω2 ⎝ ωo ⎠
Bode plots of the loop gain TL2(s) and closedloop gain GCL2(s) for positive
sequence quantities alone are plotted in Figure 27 for ωo = 2π(40) rad./s and ω2 = 2π(60)
rad./s. As seen in the figure, the loop gain plot has an infinite magnitude at frequency ω60
= 2π(60) rad./s and therefore the closedloop gain is unity with zero phase delay at this
frequency. However, contrary to the earlier case, the gaincrossover frequencies are not
displaced equally from the center frequency ω60, and the phase margin is more than 90o
at the positive slope and less than 90o at the negative slope. As a result, the closedloop
response shows peaking at the negative slope gaincrossover frequency alone giving rise
frequency response about the zero frequency and have an identical phase margin for
20 180
135
90
45
0 0
45
90
135
φm
180 3
20 3 10 100 1 .10
10 100 1 .10
Frequency (Hz) Frequency (Hz)
Figure 27 Bode plot of the loop gain TL2(s) (solid line) and closedloop gain GCL2(s) (dashed line). (Only
positive sequence response is shown.)
The complex exponential step response f2(t) of closedloop system with transfer
⎪⎧ GCL2(s) ⎪⎫
f2(t) = _ 1⎨ ⎬ (235)
⎪⎩s – jω60⎪⎭
⎡ 1
f2(t) = ejω60t ⎢1 +
2γ2
{(γ1  γ2) e(γ1 + γ2) t  (γ1 + γ2) e(γ1 − γ2) t}⎤⎥⎦ (236)
⎣
(ω2 + jω60)
where γ1 =  2 (237)
Figure 28 illustrates the complex exponential step response f2(t) for closedloop
transfer function GCL2(s) at ωo = 2π(40) rad./s and ω2 = 2π(60) rad./s. The absolute value
and argument of f2(t) are plotted in Figure 29(a) and (b), respectively. As seen in these
figures, the complex exponential step response displays an oscillatory response in both
37
magnitude and angle derivative before settling at unity magnitude and 2π(60 Hz)
frequency.
1.5
1
β
0.5
1.5
f2α(t) t
1 0 0.012 0.024 0.036 0.048 0.06
f2 0.5
0.5 1
α 1.5
1.5 1 0.5 0 0.5 1 1.5 1.5
0.5
1
1 0.5
1.5 f2β(t) t
0 0.012 0.024 0.036 0.048 0.06
0.5
1.5
Figure 28 Complex exponential step response f2(t) of the closedloop transfer function GCL2(s)
1.25 3.14
1
1.57
0.75
f2(t) ∠f2(t) t
0.5 0 0.012 0.024 0.036 0.048 0.06
1.57
0.25
t
3.14
0 0.012 0.024 0.036 0.048 0.06
(a) (b)
Figure 29 Magnitude and angle response of GCL2(s) to a complex exponential step input
A time response such as the one in Figure 28 and Figure 29 is associated with
the characteristics of peak overshoot (Pk) and settling time (Ts) that vary as a function of
frequencies ω2 and ωo. Figure 210 illustrates the results obtained for a 3% tolerance in
38
magnitude as well as angle using MathCAD® software. The phase margin was affected
0.25 ω 2/ ω o = 16
5/ ω
1.25 0.2
1.2 0.15
1.15
0.1
1.1 ω60
0.05
1.05
1 0
0 10 20 30 40 50 60 70 80 90 50 100 150 200 250 300 350 400
Figure 210 Variation of peak overshoot and settling time for vector controlled secondorder ac plants
As seen in Figure 210, the peak overshoot is affected by the phase margin (φm)
that can be easily determined from the Bode plot of loop gain TL2(s) (refer Figure 27).
The peak overshoot plot also contains as baseline that of secondorder dc systems. As
seen in this plot, the peak overshoot for secondorder ac vector controlled systems is
lesser than that for the secondorder dc systems at phase margins less than 70o. Besides
peak overshoot, the settling time for secondorder ac vector control systems is also
plotted against ωo in the same Figure 210. In this plot, however the firstorder systems
that produce ideal response with a settling time equal to 5/ωo are used as the baseline. As
seen in Figure 210, the settling time varies considerably only for values of ωo
Until now this chapter dealt with balanced threephase systems. The next section
deals with unbalanced threephase systems that are widespread throughout in power
distribution systems.
39
The stationary frame vector regulator dealt in an earlier section gives high
performance with zero steadystate error for balanced systems alone. Its transfer function
was
⎛ Ki⎞
Gc_1(s) = ⎜Kp + s ⎟ Gosc_p(s) (223)
⎝ ⎠
due to singlephase distribution circuits, and steadystate and transient imbalances in the
utility grid voltage is quite common [34,35]. Therefore, it is generally desired to regulate
quantities with zero steadystate error. Under such conditions, the controller is to be
designed to achieve a loop gain of the sum of two terms each giving an infinite gain for
ωo/2 ωo/2 ωo s
TL_2(s) = + = 2 (239)
s – jω60 s + jω60 s + ω602
⎪
and TL_2(s) ⎪⎪ → ∞ (240)
⎪s = ± jω60
Figure 211 illustrates the Bode plot of the transfer function TL_2(s). Since this
loop gain is a real transfer function its characteristics are identical for positive and
negative frequencies; and hence Bode plot is displayed here for only positive frequencies.
In this plot, ωo was chosen as 400. As seen in the figure, the infinite gain at ω60 = 2π(60)
40
rad./s implies a zero steadystate error for both the sequence quantities of 60 Hz
frequency.
The vector regulator transfer function for the plant Gp(s) to achieve the loop gain
TL_2(s) is determined as
TL_2(s)
Gc_2(s) = ^ (241)
Gp(s)
^ ^
ωo s (R + s L)
which gives Gc_2(s) = (242)
s2 + ω602
⎛ ^
ωo R⎞⎟ ⎛ 1
⎜ ^ ⎞
Gc_2(s) = ωo L + s ⎜ 2⎟
or ⎝ ⎠ (243)
⎜1 + ⎛⎜ω60⎞⎟ ⎟
⎝ ⎝ s ⎠⎠
⎛ Ki⎞
i.e., Gc_2(s) = ⎜Kp + s ⎟ Gosc(s) (244)
⎝ ⎠
ω R
^
^
where Kp = ωo L, Ki = o (224)
s
⎛ 1 ⎞
Gosc(s) = ⎜ 2⎟
and ⎜1 + ⎛⎜ω60⎞⎟ ⎟ (245)
⎝ ⎝ s ⎠⎠
60 135
40 90
Phase Angle (deg)
Magnitude (dB)
20 45
0 0
20 45
40 90
60 135
3 4 3 4
0.01 0.1 1 10 100 1 .10 1 .10 0.01 0.1 1 10 100 1 .10 1 .10
Frequency (Hz) Frequency (Hz)
The overall block diagram of the proposed stationary frame vector regulator
41
Gc_2(s) is illustrated in Figure 212. As seen in the figure, it is employed for the output
vector fout(t) to track the reference stationary frame vector fref(t) with zero steadystate
fαref(t) Ki
Kp + s
Gosc(s)
Ki
fβref(t) Kp + s
ω60 ⌠ ω60 ⌠
fβout(t) ⌡ ⌡
The vector regulator Gc_2(s) shown above is different from the P + resonant
2 Ki s
Gc_2Zm(s) = Kp + (246)
s2 + ω602
controller given by
(1 + sT)(s + α)
Gc_2Bo(s) = K (247)
s2 + ω602
However, details of controller realization and design of constant α were not provided in
[36]. The above transfer function is equivalent to the vector regulator Gc_2(s) in (244) if
42
the zero is at s = α = 0.
Thus, in order to track both positive and negative sequence quantities along with
their disturbance rejection, vector regulator Gc_2(s) has a PI cascaded with a real
coefficient oscillator Gosc(s). In the next section, the realization of proposed vector
Reference Frames
transformation [68]
e+
⎡ fd (t) ⎤ ⎡ fα(t) ⎤
⎢ e+ ⎥ = [Pe+] ⎢ ⎥ (248)
⎣ fq (t) ⎦ ⎣ fβ(t) ⎦
⎡ cos(ω60 t) sin(ω60 t) ⎤
where [Pe+] = ⎢ ⎥ (249)
⎣ sin(ω60 t) cos(ω60 t) ⎦
In terms of spacevector quantities, if fs(t) and fe+(t) denote them in stationary and
synchronous reference frames where the synchronous reference frame rotates in the anti
Suppose the complex transfer function in the stationary reference frame between
f s (t) and fs(t) is Gs(s), the equivalent transfer functions in the synchronous reference
err c
e+ e+ e– e–
frame, between vector quantities ferr(t) and f c (t), and ferr(t) and f c (t) are determined as
respectively
Ge+
( s) =
{
_ f (t)} _ {e
e+
c
=
–jω60t s
f (t)
c } = G (s + j ω
s
60) (254)
_{f (t)} _{e
e+ –jω60t s
f (t)}
err err
Ge–
_
( s) =
{f (t)} = _ {e
e–
c
jω60t s
f (t)
c } = G (s – j ω
s
and 60) (255)
_{f (t)} _{e (t)}
e– jω60t s
f
err err
e+ e– s s
and G 1 (s) and G 2 (s) as the equivalent transfer functions of G1(s) and G2(s) respectively,
‡
* denotes complex conjugate
44
s s s s
where fc1(t) and fc2(t) are the outputs of the transfer functions G1(s) and G2(s)
s
respectively, and ferr(t) is their common input quantity.
The vector regulator in the stationary reference frame can be represented in the
form of (256) as
1⎛ Ki Kp ω60⎞
Ge+ (s) = 2 ⎜Kp + s + j s ⎟ (259)
cp ⎝ ⎠
Likewise, (255) can be applied on the second term and the equivalent clockwise
1⎛ Ki Kp ω60⎞
Ge– (s) = 2 ⎜Kp + s  j s ⎟ (260)
cn ⎝ ⎠
e+ e–
Thus, the regulators Gcp(s) and Gcn(s) derived above are identical to the complex
frames that are rotating in anticlockwise and clockwise directions. Figure 213 illustrates
this realization. In the next subsection a comparative evaluation is made between the
e+
fdout(t)
e+
fderr (t)
e+
fdref (t) 1⎛ Ki⎞ e+
fdc1(t)
⎜
2⎝ Kp+
s ⎟⎠
1 ⎛Kpω60⎞
⎜ ⎟
2⎝ s ⎠
1 ⎛Kpω60⎞
⎜ ⎟
2⎝ s ⎠
e+ 1⎛ Ki⎞ e+
fqc1(t)
fqref (t) 2 ⎜⎝Kp + s ⎟⎠
e+
fqerr(t)
e+
fqout(t) (a)
e–
fdout(t)
e–
fderr (t)
e–
fdref (t) 1⎛ Ki⎞ e–
fdc1 (t)
2 ⎜⎝Kp + s ⎟⎠
1 ⎛Kpω60⎞
⎜ ⎟
2⎝ s ⎠
1 ⎛Kpω60⎞
⎜ ⎟
2⎝ s ⎠
e– 1⎛ Ki⎞ e–
fqc1 (t)
fqref (t) ⎜
2⎝ Kp+
s ⎟⎠
e–
fqerr(t)
e–
fqout(t) (b)
Figure 213 Block diagram illustrating realization of vector regulators (a) Gcp(s) in anticlockwise rotating
synchronous frame and (b) Gcn(s) in clockwise rotating synchronous frame
46
Based on the transfer function relationships between each reference frame dealt in
the earlier section, a comparison can be made between the synchronous frame and
[72], the criteria chosen as the basis for making this comparison is the number of
demerits in case the controller is applied in analog circuitry (using operational amplifiers)
as well. The final impact of such a comparative study for both analog circuitry and digital
code would be on equipment cost and planning as it evaluates the memory usage as well
Accordingly, each MAC or ITGR is considered as a unit and the total number of
such units is determined for both synchronous and stationary frame implementations. It is
to be noted that the comparison of the proposed stationary frame controller is made
against the synchronous frame implementation that was followed earlier in this section,
feedback variables that were originally in stationary reference frame into a synchronously
rotating reference frame. Every such transformation, according to (248) and (251),
regulators (refer Figure 213(a) or Figure 213(b)) contains two error determining adders,
two PI regulators and two adders that produce the final controller output, thus totaling to
six multiplyaccumulates (MACs). Further, each of the complex regulators requires four
integrators (ITGRs) to obtain the controller output along with MACs. The final controller
output in stationary reference frame is obtained after transforming back to the stationary
reference frame and adding the two regulator outputs, which takes another five MACs.
On the other hand, the stationary frame implementation contains only two PI
regulators and two oscillators (refer Figure 212). After determining the error between
the reference and feedback quantities using two MACs, the PI regulator plus oscillator
the vector regulator are tabulated in Figure 214. As seen in this figure, the total units
required for the stationary frame implementation is less than half that necessary for a
Units
Type of
Description
Implementation
MACs ITGRs Total
2 MACs f e+ (t)
out
f s (t)
out
2 MACs f e– (t) 4 0 4
out
f e+ (t)
ref
6 MACs
+
fe+(t)
c1
Synchronous 4 ITGRs
Frame f e+ (t)
out
17 8 25
5 MACs fs(t)
c
f e– (t)
ref
6 MACs
+ fe–(t)
c2
f e– (t) 4 ITGRs
out
Total 21 8 29
‡ Refer Figure 213 for details
f s (t)
ref
6 MACs
+
fs(t)
Stationary c
6 6 12
Frame s 6 ITGRs
f (t)
out
Figure 214 Comparison of synchronous frame and stationary frame implementations of vector regulator
49
2.7 Summary
In this chapter, the concepts of feedback controllers for dc quantities have been
regulators on stationary frame quantities were derived. These vector regulators have a PI
controller cascaded with an oscillator. They give a zero steadystate error when the
characteristics of peak overshoot and settling time for vector regulated systems are
evaluated against the frequencydomain parameters like the phase margin. When
transformed to synchronous rotating frames, the vector regulators are equivalent to the
evaluation of the synchronous frame and stationary frame implementations of the vector
regulator is also covered. Further comparison against various linear regulators of the
Linear Regulator
Features
Synchronous frame PI regulator [26]
ejω60t ejω60t
Zero steadystate
Ki error with
Kp + s
f s (t) fs(t) frequency
ref c
dependent transient
Ki response in
Kp + s magnitude and
phase angle [21]
ejω60t ejω60t
3.1 Introduction
pulsewidth modulation (PWM). The interface between inverter and the load/utility is
typically an LCL filter network and a transformer that are necessary for filtering and
conditioning to a utility grade voltage. However, addition of such filters increases the
order of system and may further cause resonance in the system. The inverter internal
controls described in this chapter are high speed loops aimed at achieving terminal
voltage regulation for the DR according to its generation requirements. Such controls can
also be designed to offer additional benefits of power quality conditioning and thus make
the inverter based DR a more attractive solution. In this chapter, a new approach is
presented to regulate the terminal voltage at the sensitive load bus even under load and
line imbalance conditions by applying the stationary frame vector regulator proposed in
The power circuit schematic of an inverter based DR and its filter interface
network is illustrated in Figure 31. This circuit is a particular case of that shown in
Figure 12, when the generalized impedances are balanced in all three phases and they
are transformed to the primary side of the transformer. Besides, the transformer
secondary is not connected to any other network. A block diagram of this power circuit in
the orthogonal αβ coordinate system for a balanced load Z’(s) is illustrated in Figure 32.
Its equivalent SISO representation for space vector quantities is shown in Figure 33. As
seen in Figure 33, the plant of the DR filter interface network between the inverter
output vinv(t) and load terminal voltage vload(t) is of third order since the system consists
of three dynamic energy storage elements (Lf, Cf and Lt), in addition to the dynamic
elements constituting the load (Z’). Moreover, if the plant is modified to contain
unbalanced loads in the three phases and/or multiple DRs, the block diagram in Figure
33 is unworkable.
the control output quantity is chosen to be vCf(t). The effect of load unbalance or
interconnected DR can be viewed as a disturbance input iLt(t) and the block diagram
representing the linear timeinvariant (LTI) system of the plant transfer function for space
vector quantities can be represented as shown in Figure 34. Thus, imbalances and effects
of interconnected DRs are relegated to the complex spacevector quantities, and the
54
complex transfer functions for spacevector quantities remain linear and timeinvariant.
The condition under which this assumption is valid is when the voltage controller
contains a feedforward quantity iLt(t) that makes the effective plant transfer function
unaffected by the dynamics in the energy storage element Lt, and in addition to the
Vdc
VAp VBp VCp
VNf
I'Lda
V'
I'Ldb anL
V'
I'Ldc bnL
V'cnL
I'Lda I'Ldb I'Ldc
Z'
Z'
Z'
V'nL
Figure 31 Circuit schematic of the threephase PWM inverter based DR system with equivalent load Z’
55
iLf,α(t)
vinv,α(t) 1 1 1
Lf s Cf s Z’(s)
Lt s
vCf,α(t) vload,α(t)
iLt,α(t)
iLt,β(t)
vinv,β(t) vCf,β(t)
1 1 1
Lf s Cf s Z’(s)
Lt s
vload,β(t)
iLf,β(t)
Figure 32 Block diagram of the power circuit of the DR filterinterface network for a balanced load in
orthogonal αβ coordinate system
iLf(t)
vinv(t) 1 1 1
Lf s Cf s Z’(s)
Lt s
vCf(t) vload(t)
iLt(t)
Figure 33 Block diagram of the power circuit of the filterinterface network for a balanced load for space
vector quantities
iLf(t)
vinv(t) 1 1 1
Lf s Cf s Z’(s)
Lt s
vCf(t) vload(t)
iLt(t)
Figure 34 Block diagram of the effective plant transfer function of the DR filterinterface network under
unbalanced/interconnected DR conditions for space vector quantities
The effective plant transfer function of the DR filter interface network between
vinv(t) and vCf(t) that is depicted by the block diagram in Figure 34 is of second order,
which may cause stability problems in the design of a regulator [74]. Moreover, the plant
56
response would contain beat frequencies corresponding to the excitation frequency and
the resonant frequency of the LfCf filter, unless the nonideality such as series resistance
controllers have been found to give superior response under such conditions [29,30,33].
On similar lines, for the DR filter interface, a multiloop regulator is designed with an
inner iLf(t) current loop and an outer vCf(t) voltage loop. The next section provides a
phase angle commands of vload†(t) from an external system controller and ensure an
accurate reproduction of the same at the load bus of the system, as vload(t). Furthermore,
the steady state and dynamic properties of the load and the system are different for
negative sequence and positive sequence excitations, making the design of the regulator
challenging.
adequate steady state and dynamic performance with time varying three phase sinusoidal
command signals even in the presence of unbalanced operating conditions. At the outset,
the feedback voltage regulator is built to regulate the filter capacitor voltage, vCf(t). This
enables the feedback controller to be designed around a second order system with
modulator, which provides the duty ratio commands dinv(t) for various throws of the
57
IGBT switches of the inverter, which is developed from the voltage command for the
reject any dc voltage disturbances from affecting the dynamic performance of the
realization of the spacevector modulator is typically carried out using a digital signal
processing system. At the next higher level in the controller is a current regulator that
acts as an inner feedback loop and regulates the filter inductor Lf current. In the current
loop design, capacitor voltage vCf(t) is added to the amplified error of the inductor Lf
current to decouple the voltage vCf(t) and the current iLf(t) in the plant. Since the filter
inductor current is not the ultimate quantity of interest, the steady state error performance
is not critical and hence uses a simple proportional regulator. The command current for
the inner current regulator loop itself iLf†(t) is derived from a filter capacitor Cf voltage
regulator loop. The dynamic performance and stability of the closedloop system is
improved by adding the feedforward current of inductor Lt [75]. The design of this
compensate for the voltage drop vcvm(t) across the impedance Lt representing the
interface reactor, which includes the effect of leakage inductance of the transformer by
modifying the voltage command to the capacitor voltage regulator. The modified
command is derived by appropriately multiplying the measured current with the series
Command iLt,A
Voltage
Modifier vinv,AC Vdc vinv,A iLt,B
vload,AC iLf,A
PWM To
Voltage Current
vload,BC Inverter vinv,B Transformer
Regulator Regulator and Load
iLf,B
vinv,BC vinv,C iLfB
iLfA
vCf,BC
vCf,AC
this control scheme for the inverter based DR. Stationary frame current and voltage
regulators are implemented with currents in phases A and B, and line voltages across
phases AC and BC, since two variables are adequate to describe a threephase threewire
system. For ease of implementation of sequence filters, these threephase quantities are
each of the segments in the multiloop regulator are elaborated below starting with the
The block diagram illustrating the innermost Lf current loop in the orthogonal αβ
coordinate system is shown in Figure 36. The current regulator gives the voltage
reference to the space vector modulator vinv†(t). As seen in Figure 36, a simple controller
with a constant gain Kc is employed for the purpose. The current loop has vCf(t) as a
feedforward signal to remove the dependency of current iLf(t) on voltage vCf(t). Then, the
net transfer function as seen by the current loop controller is the integral gain transfer
59
function [1/(Lf s)]. The conventional feedback controller solution for an integral gain
plant is a proportional regulator. The same solution is used for this control loop despite
the fact that it does not give a zero steadystate error for sinusoidal ac signals. Thus,
Gci(s) = Kc (31)
^ ω
Kc = L (32)
f ci
The fixed switching frequency of the PWM inverter is 4 kHz and so the
bandwidth of current loop is set below this frequency at ωci = 2π(600 Hz).
iLf(t) vCf(t)
iLf†(t) Kc vinv†(t)
(a)
iLf,α(t) vCf,α(t)
iLf,α†(t) Kc vinv,α†(t)
iLf,β†(t) Kc vinv,β†(t)
iLf,β(t) vCf,β(t)
(b)
Figure 36 Block diagram of the current regulator for (a) space vector quantities, and (b) orthogonal αβ
coordinate quantities
60
Kc
TL_i(s) = ^ (33)
Lf s
Bodeplot of the transfer function TL_i(s) is illustrated in Figure 37 for a gain of
Kc = 3.66. As seen in figure, there exists a simple pole at zero frequency indicating a zero
frequency ω60 have a steadystate error that is determined by the bandwidth ωci.
Furthermore, since the phase plot of the loop gain is confined to be within –90o, it has a
1
Gi_cl(s) =
⎛ Lf ⎞ (34)
1 + s ⎜K ⎟
⎝ c⎠
The inner current loop with closedloop transfer function Gi_cl(s) behaves as a
firstorder lowpass filter with cutoff frequency equal to the bandwidth of the current
loop ωci.
100 135
80 90
Phase Angle (deg)
Magnitude (dB)
60 45
40 0
20 45
0 90
20 135
3 4 3 4
0.01 0.1 1 10 100 1 .10 1 .10 0.01 0.1 1 10 100 1 .10 1 .10
Frequency (Hz) Frequency (Hz)
Figure 37 Bode plot of the loop gain for the current regulator loop
61
Subsequent to the design of the current feedback loop, design of the vCf(t) voltage
regulator is undertaken. By having the current iLt(t) as a feedforward signal in the vCf(t)
voltage loop design, the dependency of voltage vCf(t) on current iLt(t) is removed for
frequencies below the bandwidth of the current loop ωci. Assuming that the current iLt(t)
has no appreciable harmonic content for frequencies ω > ωci, if Gcv(s) denotes the
transfer function of the feedback regulator for voltage vCf(t), the loop gain of the voltage
regulator is determined as
1
TL_v(s) = Gcv(s) Gi_cl(s) C s (35)
f
If the dynamics of the voltage loop are limited to lower frequencies than the
bandwidth of the current loop ωci, the transfer function Gi_cl(s) can be neglected in (35).
1
Therefore, TL_v(s) ≈ Gcv(s) C s (36)
f
In the voltage vCf(t) control loop design from (36), the plant as seen by the
inductor current iLf(t) is the capacitance Cf that has an integral gain transfer function
[1/(Cf s)]. Unlike the current controller design, the voltage loop controller Gcv(s) is
designed to track with a zero steadystate error as well as reject disturbances of both
regulator for unbalanced systems that was described in the earlier chapter is used for this
purpose. As the effective plant to the voltage regulator is an integral gain transfer
function, a simple proportional gain cascaded with an oscillator would suffice. The
where Gosc(s) is the oscillator transfer function in (245). A zero integral gain (Ki) was
found sufficient here due to disturbance input feedback decoupling employed here.
Figure 38 gives the block diagram illustrating realization of filter capacitor
voltage controller Gcv(s). Such a voltage controller is equipped to regulate the threephase
vCf(t) iLt(t)
(a)
ω60 ⌠ ω60 ⌠
⌡ ⌡
vCf,α(t) iLt,α(t)
vCf,α†(t) iLf,α†(t)
Kv
† † † † † †
vCf (t) = vCf,α (t) + j vCf,β (t) iLf (t) = iLf,α (t) + j iLf,β (t)
Kv
†
vCf,β (t) iLf,β†(t)
(b)
Figure 38 Block diagram illustrating realization of filter capacitor voltage controller Gcv(s) for (a) space
vector quantities, and (b) orthogonal αβ coordinate quantities
Bode plot of the loop gain of the voltage regulator given by TL_v(s) from (35),
incorporating the controller Gcv(s) from (37) is illustrated in Figure 39 for Kv = 0.02.
63
This plot shows that fundamental frequency components @60Hz have an infinite gain.
The phase margin (φm) at the negative slope gain crossover frequency is observed to be
more than 50o thus limiting the peak overshoot to less than 5% (referring to Figure 210).
60 180
40 135
20
45
0 0
20 45
90
40
135 φm
60 180
3 4 3 4
0.01 0.1 1 10 100 1 .10 1 .10 0.01 0.1 1 10 100 1 .10 1 .10
Frequency (Hz) Frequency (Hz)
Figure 39 Bode plot of the loop gain for the voltage regulator loop
The above design procedure for the capacitor voltage regulator loop was made
assuming that there exist no components in iLt(t) above the bandwidth of the current loop
ωci. The current iLt(t), is otherwise treated as a disturbance input in the design. In order to
ascertain the disturbance iLt(t) rejection, the output impedance of the DR system both
without and with the controller is examined. The output impedance is given by the
transfer function between the capacitor voltage vCf(t) output and the reactor current iLt(t)
input. When no controller is employed for voltage regulation, the output impedance is
determined as
⎛ 1 ⎞ ⎛ 1 ⎞
Go_1(s) = ⎜C s⎟ ⎜ 1 ⎟
⎝ f ⎠ (38)
⎜1 + ⎟
⎝ Lf Cf s2⎠
From (38), it is clear that the output impedance of the LfCf filter is maximum
(infinity) at the corner frequency of the filter, viz., ω = ± 1/(LfCf) and is minimum at
the frequency ω = 0.
64
By contrast, the output impedance with the controller when vload†(t) is zero is
given by
⎪
⎛ 1 ⎞⎛ 1 ⎞
Go_2(s) ⎪⎪ =⎜ ⎟ ⎜ ⎟ (39)
⎝Cf (s + ωci)⎠ 2 Kv ωci s
⎪Vload†(s) = 0 ⎜1 + 2 2 ⎟
⎝ Cf (s + ωci) (s + ω60 )⎠
As evident from (39), the multiloop controller improves the output impedance.
± ω60. This is on account of infinite gain of Gcv(s) at ω = ± ω60. Furthermore, there exists
no infinite gain at ω = ± 1/(LfCf). Thus, the parallel resonance of the LfCf filter is
The Bode (gain) plot of output impedance without and with the controller is
illustrated in Figure 310(a) and Figure 310(b), respectively. As seen in Figure 310,
without the controller LfCf filter exhibits parallel resonance to iLt(t) at its corner
frequency of 933 Hz. By employing the proportional controller with capacitor voltage
vCf(t) feedforward in the inner current loop, this parallel resonance is damped.
Furthermore, use of the inductor current iLt(t) feedforward in the outer voltage loop
provides a low impedance for all frequencies well below the bandwidth of the current
loop ωci. The voltage regulator having singularities at ω = ±ω60 ensures a transmission
zero for 60 Hz components of iLt(t) complex space vector. This is indicated by a low gain
at the fundamental frequency in the response. Thus, the controller Gcv(s) provides good
80 80
60 60
40 40
Magnitude (dB)
Magnitude (dB)
20 20
0 0
20 20
40 40
60 60
80 80
3 4 3 4
0.01 0.1 1 10 100 1 .10 1 .10 0.01 0.1 1 10 100 1 .10 1 .10
Frequency (Hz) Frequency (Hz)
(a) (b)
Figure 310 Bode (gain) plot of the output impedance (a) without the multiloop controller and (b) with the
multiloop vector controller.
The controller thus designed, regulates the filter capacitor voltage on the primary
side of the transformer. However, the principal objective of designing a controller for the
DR system is to regulate the load bus voltage on the secondary side of the transformer
under all conditions to remain within adequate tolerance limits. For that reason, the
voltage drop vcvm(t) across transformer leakage and reactor Lt is estimated and added to
the capacitor voltage reference as it improves the load bus voltage regulation. The
expanded block diagram of the command voltage modifier block (refer Figure 35) is
illustrated in Figure 311. As seen in Figure 311, in the command voltage modifier
block, positive and negative sequence components of reactor current iLt(t) are extracted
by filters PSF and NSF, respectively. The impedance of transformer leakage together
current is essential to determine the voltage drop in the reactor and transformer leakage.
The threephase complex spacevector of this voltage drop is positive for the positive
66
sequence component in iLt(t) and negative for the negative sequence component in iLt(t),
when the impedance of the reactor and transformer leakage is assumed to be purely
inductive.
Positive
jω60 Lt sequence
filter (PSF)
iLt(t)
Negative
jω60 Lt sequence
filter (NSF)
vcvm(t)
vload†(t)
Gcv(s) iLf†(t)
vCf(t)
Figure 311 Block diagram of the command voltage modifier
Sequence filtering has been classified into two categories, viz., averaging and
delaying techniques [76]. The sequence filter proposed in this section can be categorized
earlier include single stage bandpass/bandreject filters [57,58] and cascaded sections of
complex bandpass/bandreject filters [53]. These methods are observed to give high
performance for high frequency fundamental frequencies that are used in carriersignal
injectionbased sensorless techniques [53]. On the other hand, the delaying techniques
involve combinations of the three phase quantities, with a few of those delayed by a
definite interval of time period resulting in unacceptable phase delays and complicating
Among the averaging techniques, the complex bandpass filters have been state
al [38]. Allmeling [39] has extended this transfer function for a complex bandpass filter
that passes all the chosen harmonics. The method applied by Hochgraf [77] involved
all quantities. Yuan et al [38] have described a means to perform positive sequence
⎛ 1 ⎞
GYPSF(s) = ⎜
⎛s – j ω60⎞⎟ (310)
⎜1 + ⎜ ⎟⎟
⎝ ⎝ ωo ⎠⎠
filter for the positive sequence 60 Hz quantities under unbalanced conditions, GYPSF(s)
requires a small gain ωo. Furthermore, if a similar transfer function is used to design a
negative sequence filter for a threephase system containing chiefly positive sequence
components, it allows residual positive sequence components along with phase distortion.
functions. The NSF is designed to maintain a unity gain for negative sequence
component and provide a transmission zero for the positive sequence. Thus, it passes
(with unity gain) only the 60 Hz quantity of the negative sequence and entirely eliminates
the 60 Hz quantity of positive sequence. The NSF complex transfer function is given by
⎛ 1 ⎞
GNSF(s) = ⎜
⎛2jω60 (s + j ω60)⎞⎟ (311)
⎜1 + ⎜ ⎟⎟
⎝ ⎝ ωo (s – j ω60) ⎠⎠
Bodeplots of the NSF transfer function in (311) for positive and negative
sequence components is illustrated in Figure 312. As seen in the figure, the NSF gives a
unity gain for negative sequence components at frequencies near 60 Hz, while
20 20
0 0
Magnitude (dB)
Magnitude (dB)
20 20
40 40
60 60
80 80
100 100
4 3 3 4
1 .10 1 .10 100 10 1 0.1 0.01 0.01 0.1 1 10 100 1 .10 1 .10
Frequency (Hz) Frequency (Hz)
135 135
90 90
Phase Angle (deg)
45 45
0 0
45 45
90 90
135 135
4 3 3 4
1 .10 1 .10 100 10 1 0.1 0.01 0.01 0.1 1 10 100 1 .10 1 .10
Frequency (Hz) Frequency (Hz)
(a) (b)
Figure 312 Bode plot of the transfer function GNSF(s) for (a) negative sequence and (b) positive sequence
components of the space vector
Gfn(s)
GNSF(s) = 1 + G (s) , (312)
fn
the transfer function GNSF(s) can be realized as a unity gain feedback controller of
⎛ ωo (s – j ω60) ⎞
Gfn(s) = ⎜ ⎟ (313)
⎝2jω60 (s + j ω60)⎠
⎛ ωo ⎞ ⎪⎧⎨ω60 ⎪⎫⎬ ⎛ 1 ⎞
Gfn(s) = ⎜ ⎟ +j ⎜
⎝2ω60⎠ ⎩⎪ s ⎭⎪ jω ⎟ (314)
⎜1 + 60⎟
⎝ s ⎠
The equation (314) consists of three factors: (i) the first factor, (ωo/(2ω60)) is a
constant gain; (ii) the second factor is a complex gain {(ω60/s) + j} of the forward
coupling form (Refer Figure 313); and (iii) the third factor is the oscillator Gosc_n(s) in
(226) that provides forced oscillations for negative sequence components while
neglecting positive sequence components. The negative sequence filter, GNSF(s), may be
realized in αβ coordinates as in Figure 313, with the three factors clearly outlined.
iLt,α(t) iLt,αn(t)
ωo/(2ω60) ω60 ⌠
⌡
ω60 ⌠
iLt(t) = ⌡
(i) (ii) (iii) iLt,n(t) =
iLt,α(t) + j iLt,β(t)
iLt,αn(t) + j iLt,βn(t)
ω60 ⌠
⌡
iLt,β(t) iLt,βn(t)
ωo/(2ω60) ω60 ⌠
⌡
Figure 313 Block diagram illustrating realization of negative sequence filter (NSF) in the command
voltage modifier
70
The loop gain of the NSF is equal to Gfn(s), due to the unity gain feedback. The
Bode plot of the openloop gain of the NSF for ωo = 200 is shown in Figure 314(a) and
Figure 314(b) for negative and positive sequence quantities of the complex space vector
iLt(t).
As seen in Figure 314, the dashed lines indicating the negative sequence
response show an infinite gain at ω = ω60; and the solid lines indicating the positive
sequence response show a transmission zero at ω = +ω60. Therefore, the NSF allows only
the negative sequence components of iLt(t) to pass with a unity gain while blocking the
80 80
60 60
40 40
Magnitude (dB)
Magnitude (dB)
20 20
0 0
20 20
40 40
60 60
80 80
100 100
120 120
4 3 3 4
1 .10 1 .10 100 10 1 0.1 0.01 0.01 0.1 1 10 100 1 .10 1 .10
Frequency (Hz) Frequency (Hz)
135 135
90 90
Phase Angle (deg)
45 45
0 0
45 45
90 90
135 135
4 3 3 4
1 .10 1 .10 100 10 1 0.1 0.01 0.01 0.1 1 10 100 1 .10 1 .10
Frequency (Hz) Frequency (Hz)
(a) (b)
Figure 314 Bode plot of the loop gain for the negative sequence filter (NSF) Gfn(s) for (a) negative
sequence and (b) positive sequence components of the space vector iLt(t)
In the NSF shown in Figure 313, the negative sequence component of the current
iLt(t) is iLt,n(t). The negative sequence current iLt,n(t) is multiplied by a complex gain ‘
71
jω60 Lt’ to determine the negative sequence component of vcvm(t) as shown in Figure 315
Figure 315 Block diagram illustrating realization of complex gain ‘jω60 Lt’ for the output of the NSF in
αβ coordinates
Similar to the NSF, the positive sequence filter (PSF) is designed to maintain
unity gain for the positive sequence component while providing a transmission zero for
the negative sequence component in iLt(t). The PSF complex transfer function is given by
⎛ 1 ⎞
GPSF(s) = ⎜
⎛2jω (s – j ω60)⎞⎟ (315)
⎜1 + ⎜ 60 ⎟⎟
⎝ ⎝ ωo (s + j ω60) ⎠⎠
The PSF can be realized by the NSF block diagram in Figure 313 by swapping
the α and β terminals at both the input and output. Thereafter, the positive sequence
component of the current iLt,p(t) thus extracted is multiplied by a complex gain ‘jω60 Lt’
command voltage modifier output vcvm(t) is the sum of the estimated positive and
negative sequence voltage drops across the leakage inductance of the transformer plus the
The transient response of the sequence filter proposed in the earlier section is
related to the gain ωo. Figure 316 gives the Bode plot of the negative sequence filter for
values of gain ωo = 20, 200 and 2000. As seen in figure, when ωo is increased the width
of the passband is also increased but that of the stopband is decreased. The transient
response of the negative sequence quantities is influenced by the passband width while
that of the positive sequence by the stopband width of the frequency response. Thus, the
unity gain and zero phase lag for the negative sequence and a total elimination of positive
a closeloop system, the phase lag/lead introduced by it in the loop gain transfer function
can be reduced by increasing the gain ωo for stabilizing the closedloop system against a
degradation of only the transient performance of the positive sequence stopband width.
73
20 20
Magnitude (dB) 0 0
Magnitude (dB)
20 20
40 40
60 ωo ωo 60 ωo ωo
80 80
100 100
3 4
4 3
1 .10 1 .10 100 10 1 0.1 0.01 0.01 0.1 1 10 100 1 .10 1 .10
Frequency (Hz) Frequency (Hz)
135 135
90 90
Phase Angle (deg)
necessary to establish the stability of such a filter. In this section, stability of one of the
two proposed filters, viz., the negative sequence filter (NSF) is investigated. While the
criterion [78] has been exhaustively dealt in several textbooks on control systems [71,79],
have applied the Nyquist diagram on complexcoefficient transfer functions for three
phase power systems. Extending this approach, guidelines to ensure the stability of such
transfer functions are presented here. For an interpretation of the stability of complex
transfer functions with regard to the multiinput multioutput (MIMO) systems theory,
The vector block diagram of NSF realization explicitly showing the feedback
loop is illustrated in Figure 317. As seen in the figure, the loop gain TL(s) is also the
forward gain Gfn(s) of the unity feedback NSF system and it is of the form
⎛M (s – j ω60)⎞
TL(s) = ⎜ ⎟ (317)
⎝ (s + j ω60) ⎠
f(t) fn(t)
Gfn(s)
Figure 317 Block diagram illustrating realization of negative sequence filter (NSF) as a unity feedback
system
where M is a constant in the complex plane C. Two cases of complex constant M are
ωo ωo ⎛s – j ω60 ⎞
Case (i): M = : TL(s) = ⎜ ⎟
2ω60 2ω60 ⎝s + j ω60⎠
The locus of the Nyquist path in the splane and the TL plane corresponding to
this case are displayed in Figure 318. The transfer function TL(s) has a pole at s = jω60
and a zero at s = jω60. Since there exists a singularity of TL(s) on the jω axis at s = jω60,
75
the Nyquist path is modified to pass along a semicircle of infinitesimal radius around s =
jω60. The Nyquist path is divided into 5 segments and the corresponding segments in the
TL locus are marked with the same numbers. It is observed that unlike realcoefficient
transfer functions, the TL locus is not symmetric for positive and negative frequencies for
jω Im{TL}
splane TLplane
j∞
5
jω60 s = j∞
1 1
jω60 ∞ σ (1,0) Re{TL}
2 4 3
3 5
4 2
j∞
s = jω60
P=0 N≠0
Figure 318 Nyquist path on the splane and the TL locus for M = ωo/(2ω60)
As seen in Figure 318, while the number of poles of TL(s) in the right half of the
N ≠ 0. This is because the TL locus passes through the point (1,0). Therefore, the number
the Nyquist stability criterion the feedback system with this loop gain is not stable.
ωo ⎛ s – jω60 ⎞
Verification: 1 + TL(s) = 1 + ⎜ ⎟ , which has a zero on the imaginary axis and
2ω60 ⎝s + jω60⎠
In this case, the loop gain transfer function in Case (i) is multiplied by the
imaginary operator ‘j’. The Nyquist path in the splane and the TL locus corresponding to
this case are displayed in Figure 319. As seen in the figure, multiplying the loop gain by
‘j’ has resulted in the TL locus to be rotated anticlockwise by 90o as compared to that in
Case (i). While the number of poles of TL(s) in the right half of the splane P = 0, the
number of encirclements of the TL locus about (1,0) in this case is N = 0. Therefore, the
number of zeros of 1 + TL(s) in the right half of the splane Z = N + P = 0. Hence, the
jω Im{TL}
splane TLplane
j∞ s = −j∞ 3
5 s = jω60
jω60 4
1 s = j∞
5
P=0 N=0
Figure 319 Nyquist path on the splane and the TL locus for M = jωo/(2ω60)
jωo ⎛s – jω60 ⎞
Verification: 1 + TL(s) = 1 + ⎜ ⎟, which has a zero in the left half of
2ω60 ⎝s + jω60⎠
the phase of loop gain TL(s) with the complex coefficient M. However, it is not always
coefficient transfer functions. If the number of poles of TL(s) in the right half of the s
77
plane P = 0, it is also possible to evaluate stability on the basis of Bode plots [71,82]. The
Bode plots of loop gain of TL(s) for Cases (i) and (ii) are displayed in Figure 320 and
Figure 321, respectively. These plots have log sweep of frequency as it would provide
the magnitude and phase variation over a wide frequency range that is beneficial for
visualizing broad frequency phenomena in power systems. The left section of each Bode
plot represents response for negative frequencies, and right section that for positive
frequencies of space vector. As noted earlier, the frequency response for positive
sequence quantities is indicated by the characteristics for positive frequencies, and those
As seen in Figure 318, the Nyquist diagram for Case (i) M = ωo/(2ω60) showed
instability as the TL locus passed through (1,0) on the negative real axis. The
corresponding Bode plot of loop gain TL(s) is displayed in Figure 320. In the Bode
magnitude plot, the phase angle at the gain crossover frequency is +180o. Hence, the
On the other hand, the Nyquist diagram for Case (ii) M = jωo/(2ω60) in Figure
319 showed that the TL locus does not encircle the point (1,0) and therefore the
feedback system is stable. This is indicated in the Bode plot of Figure 321 by a 90o
From the stability analysis presented in this subsection, it has been proved that the
proposed negative sequence filter is stable as its forward gain Gfn(s) has a loop gain TL(s)
80 80
60 60
Magnitude (dB)
Magnitude (dB)
40 40
20 20
0 0
20 20
40 40
60 60
80 80
3 2 1 0 0 1 2 3
10 10 10 10 10 10 10 10
Frequency (Hz) Frequency (Hz)
180 180
135 135
Phase Angle (deg)
80 80
60 60
Magnitude (dB)
Magnitude (dB)
40 40
20 20
0 0
20 20
40 40
60 60
80 80
3 2 1 0
10 10 10 10 0 1 2 3
Frequency (Hz) 10 10 10 10
Frequency (Hz)
180 180
135 135
Phase Angle (deg)
Phase Angle (deg)
90 90
45 45
0 0
45 45
90 90
135 135
180 180
3 2 1 0 0 1 2 3
10 10 10 10 10 10 10 10
Frequency (Hz) Frequency (Hz)
(a) (b)
Figure 321 Bode plot of loop gain TL(s) for M = jωo/(2ω60) for (a) negative and (b) positive sequence
components of the space vector
79
This section presents selected results from simulation studies conducted on the
performance of the controllers designed for voltage regulation under imbalance are
presented. The various controllers tested include filter inductor current regulator, filter
capacitor voltage regulator, and the command voltage modifier containing the positive
and negative sequence filters. Digital simulation has been carried out in Matlab®
has been utilized for creation of various power quality events in the simulation.
The simulation model of the DR system assumes an averaged model of the pulse
width modulator and the inverter together with dc bus voltage feedforward (refer Figure
35). The physical plant consisting of the DR filter interface network is modeled by linear
ODEs describing the threephase circuits. The stationary frame current and voltage
regulators are implemented with currents in phases A and C, and line voltages across
phases AB and CB, since two variables are adequate to describe a threephase threewire
system. Voltage imbalances at the load terminal are caused by load imbalances in the
three phases due to unequal rms currents drawn from each phase of the voltage source.
The perunit phasor plots in Figure 322 portray a typical case of unequal singlephase
unregulated.
80
1.5 1
0.95∠31
1
0.92∠153 0.5
0.5 0.95∠88
1.5 1 0.5 0 0.5 1 1.5 1 0.5 0 0.5 1
0.67∠76 0.5
0.5
1.3∠162 1
0.83∠43 1.5 1
ILa VanL
ILb VbnL
ILc VcnL
Figure 322 Effect of load imbalance (left) on the unregulated terminal voltage (right)
The effect of voltage imbalances or sags is a variation in the reactor current iLt.
Figure 323 illustrates the performance of the voltage regulator for an incremental
step change in voltage command from 70% to 100% of the rated voltage of 480 VRMS.
Waveforms displayed in Figure 323 are the αβ components of reference voltage, actual
voltage that tracks the reference and the error between the reference and actual values. As
seen in the figure, the capacitor voltage response reaches close to the steady state in about
one and half 60 Hz cycles. During this transient, the actual voltage of filter capacitor
800
α variables
400
400
800
0.3 0.32 0.34 0.36 0.38 0.4 0.42 0.44 0.46
800
β variables
400
400
800
0.3 0.32 0.34 0.36 0.38 0.4 0.42 0.44 0.46
Time (s)
Figure 323 Operation of the voltage control loop for vCf with the response vCf for a complex exponential
step input voltage reference vCf†.
2000
1500
1000
vCf,AC (V)
500
500
1000
1500
2000
0.1 0.2 0.3 0.4 0.5
Time (s)
Figure 324 Simulation waveform of line voltage vCf,AC illustrating the response for a step change in
voltage reference vCf† from 70% to 100% of rated voltage of 480 VRMS
The first test in this case involves a threephase 10 kW resistive load. Figure 325
illustrates the steadystate voltage waveforms at the load terminals across phases ac and
vac(V)
0
200
400
0.5 0.52 0.54 0.56 0.58 0.6 0.62 0.64 0.66 0.68 0.7
400
200
vbc(V)
0
200
400
0.5 0.52 0.54 0.56 0.58 0.6 0.62 0.64 0.66 0.68 0.7
80
40
ia(A)
0
40
80
0.5 0.52 0.54 0.56 0.58 0.6 0.62 0.64 0.66 0.68 0.7
80
40
0
ic(A)
40
80
0.5 0.52 0.54 0.56 0.58 0.6 0.62 0.64 0.66 0.68 0.7
Time (s)
Figure 325 Load terminal lineline voltage ac & bc (top two) and load current in a & b phases (bottom
two) for a balanced 10 kW resistive load.
Another test carried out to observe the dynamic response of the controller under
balanced conditions involved abrupt load change. Figure 326 demonstrates the response
load was abruptly connected using an threepole singlethrow switch. As seen in Figure
326, load terminal voltages on the secondary side of 480 VΔ : 208 (120 VY)
transformer respond fairly well with minor glitches at the instant of noload to load
switching. If it is assumed that the load is switched ON at t = tON, as the current is zero at
t = tON– the voltage is also zero; and as the ac current begins to flow through the load at t
= tON+ the voltage at load terminals regains its sinusoidal form. This voltage recovery
vac(V)
0
200
400
0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.14
400
200
vbc(V)
0
200
400
0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.14
20
10
ia(A)
0
10
20
0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.14
20
10
0
ic(A)
10
20
0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.14
Time (s)
Figure 326 Simulation waveforms illustrating DR operation during transition from noload to a 2 kW
balanced threephase resistive. Load terminal lineline voltage ac & bc (top two) and load current in a & b
phases (bottom two).
noload is illustrated Figure 327. As seen in this figure, the DR responds satisfactorily
400
200
vac(V)
0
200
400
0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.14
400
200
vbc(V)
0
200
400
0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.14
20
10
ia(A)
0
10
20
0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.14
20
10
ic(A)
0
10
20
0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.14
Time (s)
Figure 327 Simulation waveforms illustrating DR operation during transition from a 2 kW balanced three
phase resistive load to noload. Load terminal lineline voltage ac & bc (top two) and load current in a & b
phases (bottom two).
84
shown to satisfactorily regulate the terminal voltage under balanced load conditions. The
next case involves tests under an unbalanced threephase loading of the DR.
Inorder to test the performance of the controllers under unbalanced conditions, the
steadystate voltage waveforms at the load terminals across phases ac and bc as well as
line currents through phases a and b. Table 31 provides a comparison of the regulated
load terminal voltage with that of unregulated case. As seen in Table 31, the lineline
voltages are regulated fairly close to 208 V in all three phases. Small inconsistencies
400
200
vac(V)
0
200
400
0.9 0.92 0.94 0.96 0.98 1 1.02 1.04 1.06 1.08 1.1
400
200
vbc(V)
0
200
400
0.9 0.92 0.94 0.96 0.98 1 1.02 1.04 1.06 1.08 1.1
80
40
ia(A)
0
40
80
0.9 0.92 0.94 0.96 0.98 1 1.02 1.04 1.06 1.08 1.1
80
40
ic(A)
0
40
80
0.9 0.92 0.94 0.96 0.98 1 1.02 1.04 1.06 1.08 1.1
Time (s)
Figure 328 Simulation steadystate waveforms illustrating the load terminal voltage and load current for
an unbalanced load
85
consisting of sequence filters PSF and NSF has a major impact on the effectiveness of
regulation of load terminal voltage. The functioning of PSF and NSF are demonstrated in
Figure 329. It illustrates the filter input which is the transformer primary side current,
40
20
Input
0
20
40
0.96 0.98 1 1.02 1.04 1.06 1.08 1.1 1.12
40
PSF output
20
0
20
40
0.96 0.98 1 1.02 1.04 1.06 1.08 1.1 1.12
40
NSF output
20
0
20
40
0.96 0.98 1 1.02 1.04 1.06 1.08 1.1 1.12
Time (s)
Figure 329 α and β components of transformer primary current, PSF output and NSF output waveforms
for an unbalanced load
As seen in Figure 614, the PSF and NSF satisfactorily extract the positive and
negative sequence components thus aiding in accurate compensation of the reactor and
3.6 Summary
regulator is applied to inverter based distributed resources to regulate the load terminal
voltage with zero steadystate error even under unbalanced conditions. The multiloop
regulator consists of three segments – inner current regulator for the filter inductor
current, outer voltage vector regulator for the filter capacitor voltage, and the command
voltage modifier to compensate for the transformer leakage impedance voltage drop.
Frequency domain analysis techniques in the form of Bode plots for individual sequence
Further, by making use of complex transfer functions, novel negative and positive
bandreject sections are proposed. The stability of these filters has been explained using
Nyquist diagrams and Bode plots. Table 32 gives a comparison of their frequency
response characteristics against the stateoftheart synchronous frame lowpass filter and
cascaded filters. As seen in Table 32, the synchronous frame negative sequence filter for
cascaded filter results in significant phase distortion of its output. The proposed
frequency/sequence selective complex filter gives high quality output without any phase
distortion. On the downside, its response to passband frequency components and stop
band frequency components is complementary in nature. For a very high gain ωo in the
negative sequence filter, while the time response for passing negative sequence 60 Hz
components is greatly improved, the time response for rejecting positive sequence 60 Hz
87
effects of power quality phenomena at the sensitive load bus. In this chapter, the vector
ac systems.
20
ejω60t
jω60t
e
40
180 120 60 0 60 120 180
1 Frequency (Hz)
45
0
45
90
135
180 120 60 0 60 120 180
Frequency (Hz)
88
Magnitude (dB)
20
ejω60t e2jω60t 40
60
80
s 1
f (t) 100
in 1 + s/ωo 180 120 60 0 60 120 180
Frequency (Hz)
135
90
20
40
60
80
1 100
⎛ ⎞ 180 120 60 0 60 120 180
s
f (t) ⎜ ⎛ 2j ω60 (s + j ω60 ⎞
) ⎟ f (t)s
Frequency (Hz)
in ⎜1 + ⎜ ⎟⎟ out
⎝ ⎝ ωo (s – j ω60) ⎠⎠ 135
90
Phase Angle (deg)
45
0
45
90
135
180 120 60 0 60 120 180
Frequency (Hz)
89
In the previous chapter, the design of a multiloop regulator for control of the load
Such a DR controller requires the ac voltage vector as input. The reference voltage space
vector can be set according to the nominal system specifications under standalone
phase of voltage vector such that DRs can be dispatched to share the overall load in a
predetermined manner.
This chapter deals with the generation control of the DR which provides the
magnitude and phase information to the multiloop regulator. The generation control
involves active and reactive power controllers that allow parallel operation of DRs in the
microgrid. The dynamic (smallsignal) behavior analysis for microgrids having a chain
topology is also covered in this chapter. Both real power and reactive power flows in the
microgrid are considered. Guidelines are provided for design of the active power
frequency and reactive powervoltage controls on the basis of IEEE P1547 performance
specifications [18]. Finally, selected results from simulation of DRs equipped with these
4.1 Introduction
inverter with LCL filter and a transformer is illustrated in Figure 41. Several such DR
DR is operated to regulate the load bus voltage to a value within the tolerance limits. For
that reason, the DR along with the LC filter and transformer is approximated by a voltage
source as illustrated in the figure while conducting theoretical analysis on power studies.
This assumption is especially valid since the external power generation controls of the
inverter are designed to have a bandwidth lower by atleast one order as compared to the
conducted by [60,64] have shown that the generation control would interact with the
internal controls if their bandwidths are not far apart. The power studies in the microgrid
are conducted with all variables in p.u.. It is to be noted that the regulated voltage on both
the high and lowvoltage sides of the transformer would be the same in p.u. after
comprises the magnitude (V) and frequency (ω60) information. Such a specification
constitutes the voltage command that needs to be provided to each inverter based DR in
the microgrid. A block diagram illustrating the generation of voltage command for each
inverter based DR is shown in Figure 42. As seen in Figure 42, the generation
based on the real and reactive powers drawn from the DR. These incremental variables
91
are added to their nominal values to determine the voltage command vector to the
inverter. The instantaneous real and reactive powers [50] at the DR terminal, denoted as
PL and QL, respectively, can be easily computed from the measured quantities vt and iLtp,
which is the positive sequence component of iLt. By the use of the positive sequence filter
output iLtp instead of iLt, it is ensured that the power measurement does not contain ac
ripple under unbalanced conditions. The following section in this chapter provides further
vCf
DSP
Controls
vt = V ejω60t
PL,
QL
Figure 41 Voltage regulated DR’s analogy to a voltage source for power studies
92
Vo
ΔV
PL V
†
Generation
Controller
Δω †
QL ω60
ωo
large rotating machines has been well established and reported in various literature
[61,62,64]. The frequency droop features naturally offered by the components of the
system against an increase in load support paralleling of generators and sharing of the
active power. Figure 43 demonstrates in deviation values, indicated by a “Δ”, how a
change in load demand (ΔPL) causes a deviation in speed/frequency (Δω) for a large
rotating machine based synchronous generator [61]. The generator modeling is made by
means of the relationships between deviations in mechanical power ΔPM and speed Δω.
The input labeled ‘ΔPLref’ is the load ref. setpoint that is the control input to shift the
generator’s governor characteristic in order to give the reference frequency at any desired
power output. Frequency droop follows after a change in load demand had been initially
93
provided by the inertia of the rotating mass and there is a delay in the response of the
primemover to accommodate the change in load. Governors are employed to allow the
controller to detect the change in the machine speed and thereby vary the valve input of
the primemover. There are two forms of governing in a power system, viz., generator
governing and load governing [62]. When a load increase causes a decrease in machine
speed, there is a decrease in the supply frequency, which eventually results in a net
decrease in the load. This selfgoverning attribute of a rotating machine generator is the
principal basis that makes possible paralleling of generators.While the load governing
Prime
mover ΔPL D
Speed governing
Figure 43 Block diagram of the generation controller for a rotating machine ‡
As seen in Figure 43, the speed governor for generation control is modeled by a
gain ρ for the speed sensor and a firstorder lag with timeconstant TG. The primemover
order lag transfer function with timeconstant TCH (known as charging timeconstant) that
‡
Source: [61]
94
signifies the delay in the production of the mechanical power output after change in the
valve position. The effect of generator inertia is modeled by the angular momentum of
the machine M and the influence of load governing by the frequency dependency factor
on the physical device characteristics such as the governor, primemover, machine inertia
electronic inverters offer a good deal of flexibility in the design of control parameters. A
key difference between a microgrid consisting of inverter based DRs and a large power
system with rotating machines is that the DRs do not possess inertia as they contain
microsources such as fuel cells or microturbines and power electronic converters. The
inertia in a rotating machine aids in supplying transient load support and further offers a
speed droop that is used by the speed governor to change the power input to the prime
mover in response to the change in load demand. In a DR, the energy storage on the dc
link of the inverter can provide support in meeting the transient load demands until the
output.
distributed UPS was proposed by Chandorkar [60]. This controller is illustrated in the
block diagram shown in Figure 44 and applied for the active power frequency control of
the DR. It has a frequency droop with a proportional gain transfer function (b > 0), which
provides the necessary load governing functionality that is beneficial for paralleling DR
units. The active power controller includes a frequency restoration loop whose function is
analogous to the speed governor in a rotating machine generator (refer Figure 43). A
notable difference from the controller for rotating machine is that the DR controller does
not have a firstorder lag transfer function corresponding to that of the primemover with
a charging timeconstant.
1/β
ΔPLref 1 ΔPR Δω
1 + s/ωG b
ΔPL
Frequency droop
Frequency restoration
for DR in Figure 44 is illustrated in Figure 45. The parameters in Figure 45 are
⎡ b/β ⎤
Kbβ = ⎢ ⎥ (41)
⎣1 + b/β⎦
‡
adapted from the distributed UPS controller proposed by Chandorkar [60]
96
in a DR upon load transient, and 0 < Kbβ < 1 so that a very low value Kbβ → 0 denotes
From the smallsignal model block diagram in Figure 45, the transfer function
between the deviation in frequency Δω and deviation in load demand ΔPL alone is a lead
lag filter
Δω(s) ⎡1 + s/ωG⎤
= b (1  Kbβ) ⎢ ⎥ (43)
ΔPL(s) ⎣ 1 + s/ωb ⎦
Δω(s) ⎡ 1 ⎤
= b (1  Kbβ) ⎢ ⎥ (44)
ΔPLref(s) ⎣1 + s/ωb⎦
1 Δω
PLref 1 + s/ωG b
1
PL Kbβ Δωh
1 + s/ωb
1 Δω
ΔPLref 1 + s/ωG b
1
ΔPL Kbβ Δωh
1 + s/ωb
The two transfer functions show a firstorder time response of time constant 1/ωb
97
and a final steadystate deviation in frequency of b(1  Kbβ). The steadystate frequency
droop can be graphically represented as shown in Figure 46(a) and Figure 46(b) for the
incremental variables and actual variables, respectively. As seen in these figures, the
straight line can be shifted along the ΔPL (PL) axis by varying ΔPLref (PLref). The rated
Δω ω
PL_ref < PLR
ΔPLref = 0
Slope =  b Slope =  b
ΔPL ωo
Slope = Slope =
 b(1Kbβ)  b(1Kbβ)
PL
0 1.0 p.u.
(a) (b)
Figure 46 Steadystate voltage droop of a DR in the standalone mode of operation. (a) Deviation
quantities (b) actual quantities.
converters, their current capacity is limited by the device ratings. It is therefore necessary
to control the apparent power drawn from a DR in a microgrid. This can be achieved by
having a controller for regulating reactive power drawn from the DR along with the
It has been proved beyond doubt through power flow studies conducted on an
references for the DR can cause excessive reactive power circulation between the
98
generators [63]. This fact has been recognized earlier in the paralleling of synchronous
exchange between the two generators due to discrepancies in the voltage setpoints, it is
essential to have soft setpoints in voltage magnitude. The proposed reactive power
voltage controller achieves this task – a reactive load increase causes a decrease in
voltage magnitude setpoint that eventually results in a net decrease in the reactive load.
supplied by the DR on the voltage magnitude at the load bus. Figure 47 shows a block
diagram of the proposed reactive power controller. As seen in this figure, a simple
feedback controller containing firstorder lag is employed for the deviation in the reactive
controller proposed here does not contain a voltage restoring loop. Moreover, it does not
achieve a zero steadystate error but provides a voltage droop upon an increase in the
reactive load. As seen in Figure 47, the input labeled ‘ΔQLref’ is the load ref. setpoint
that is the control input to shift the DR’s voltage regulator characteristic in order to give
1 1 ΔV
QLref Dq 1 + s/ωq
QL
Voltage droop
1 1 ΔV
ΔQLref
Dq 1 + s/ωq
ΔQL
Voltage droop
Figure 47 Block diagram of the proposed reactive power controller for a DR
Voltage droop is a firstorder lag function and not instantaneous. The active
modulation control, which varies the modulation index according to the dc bus voltage
variations in the inverter of the DR, assists in providing a fixed voltage magnitude
against load variation for some duration before other controls take precedence.
voltage droop at a particular bus by voltage droop curves. In order to generate the voltage
droop characteristics of the DR with the proposed controller, the relation between the
deviation in voltage and the change in reactive load is reexamined. This relation is the
sum of the response of ΔV to ΔQL and ΔQLref when acting alone, and is determined as
⎛ 1 ⎞⎛ 1 ⎞
ΔV(s) = ⎜D ⎟⎜ ⎟ (ΔQLref(s) – ΔQL(s)) (45)
⎝ q⎠⎝1 + s/ωq⎠
ΔQLref – ΔQL
ΔV = Dq (46)
shown in Figure 48(a). As seen in the figure, the straight line can be shifted along the
ΔQL axis by varying the load ref. setpoint. The same curves are displayed in Figure
48(b) for the actual voltage V and power demand QL variables. Representing the
nominal values of the quantities by a subscript “o”, the actual quantities are determined as
and V = Vo + ΔV (48)
Thus, Figure 48(b) characterizes the steadystate voltage droop feature of the
DR. As seen in Figure 48(b), the actual quantity corresponding to the load ref. setpoint
is QL_ref. The value of QL_ref corresponding to a load ref. setpoint of zero is QL_ref = 0. A
positive value of load ref. setpoint would refer to QL_ref > 0 and a negative value to QL_ref
< 0. By changing QL_ref the controller can be set to give nominal voltage at any desired
reactive load condition. According to the convention employed, the load is inductive
when it draws a positive reactive power, and capacitive when it draws a negative reactive
power.
101
V
ΔV QL_ref = 0
Slope =
Dq1 QL_ref > 0
ΔQLref > 0 Vo
ΔQL Slope =
Dq1
QL_ref < 0
ΔQLref < 0
ΔQLref = 0 QL
0 1.0 p.u.
1.0 p.u.
(a) (b)
Figure 48 Steadystate voltage droop of a standalone DR. (a) Deviation quantities (b) actual quantities.
controller presented in this section along with the active powerfrequency controller. The
following section.
these controllers are programmed in a digital signal processor, their parameters need to
be designed to meet conditions that guaranty stable decentralized operation when DRs
are connected in wide varying microgrid configurations. In this section, a set of design
constraints are developed for the DR controller gains that would guaranty stable
operation of the microgrid. The following assumptions are made in the analysis of
(ii) Inverter based DRs operate within their maximum capacity limits
(iii) Dynamics of the inverter internal controls are fast as compared to the
(iv) Tielines between any two sources in the microgrid are purely inductive
in nature
Consider the operation of a single DR when connected to the grid supply through
an inductive tieline as shown in Figure 49(a). Grid supply generally has enormous
infinite bus that is characterized by a stiff frequency. Therefore, any load variation at the
grid supply end does not cause a droop in its frequency in either transient or steadystate.
The smallsignal statevariable schematic of this system under zero input conditions is
given in Figure 49(b). While the real power loop is depicted in the right half of this
PL1, QL1
X1,g
Infinite Bus
V1∠δ1 E∠0o
DR1
(a) Singleline diagram
ΔV1(s) Δω 1(s)
ωq1
1/Dq1 b1
s
ω b1
Kbβ1
s
Dqtie1,g
Δωh1(s)
Δδ1(s) 1
Po1,g
s
(b) Smallsignal statevariable schematic under zero input conditions
d
dt Δωh1 =  ωb1 Δωh1  b1 b1/β1 ωG1 (Po1g Δδ1g) , (49)
d
dt Δδ1g =  Δωh1  b1 (Po1g Δδ1g) (410)
d ⎛⎛ Dqtie1g⎞ ⎞
and dt ΔV1 =  ωq1 ⎜⎝⎜⎝1 + Dq1 ⎟⎠ ΔV1⎟⎠ (411)
V1o E 2V1o  E
where Po1g = X and Dqtie1g = X1g (412)
1g
d
dt x = A x + B u (413)
Of primary interest is the system matrix, A, from which the system stability can
⎡ ω1
b1 b1(b1/β1)ωG1Po1g 0
⎤
A=
⎢ b1Po1g 0 ⎥ (415)
⎢ 0 Dqtie1g
ωq1 ⎛1 + D ⎞⎥
⎣ 0
⎝ q1 ⎠⎦
From eigenvalue studies on matrix A, it has been observed that the single grid
interfaced DR characterized by the above system matrix has real distinct eigenvalues and
is stable if the constants b1, β1, ωG1, ωq1, Dq1 are all positive and the maximum droop in
For the state vector xT = [Δωh1 Δδ12 Δωh2 ΔV1 ΔV2], the system matrix in the case of
⎡ 1 ⎤
ωb1 b1(b1/β1)ωG1Po12 0 0 0
⎢0 ⎥
(b1+b2)Po12 1 0 0
b2(b2/β2)ωG2Po12 ωb2 0 0
A=
⎢0 0 0
Dqtie12
ωq1 ⎛1 + D ⎞
ωq1V1o ⎥ (416)
⎢ ⎝ q1 ⎠
⎥
Dq1X12
⎣0 0 0
ωq2V2o
Dq2X12 ⎝
Dqtie21
ωq2 ⎛1 + D
q2 ⎠⎦
⎞
V1∠δ1 V2∠δ2
DR1 DR2
(a) Singleline diagram
Dqtie1,2 ω b1
Kbβ1
s
V1o
X1,2 Δω h1(s)
V2o
Δδ 1,2(s) 1
X1,2 Po1,2
s
Dqtie2,1
ΔV2(s) ω q2
1/Dq2 b2
s Δω 2(s)
ωb2
Kbβ2
s
Δω h2(s)
In the case of a twoDR system also, it has been observed through eigenvalue
studies that the system matrix has real distinct eigenvalues and is stable if the constants
(k = 1, 2) bk, βk, ωGk, ωqk, Dqk are all positive and the maximum droop in voltage is less
than the nominal voltage. On similar lines, system matrices can be constructed for
Cardell and Tabors [67] have shown that microgrids that were stable with only
few combustion turbine DRs can become unstable when more combustion turbines were
106
added in the system. They further suggested that frequency stability problems may arise
the active powerfrequency control flow section have been calculated for typical control
gains in systems of order 3 through 7 etc. Figure 411 illustrates the dominant
eigenvalues as the order of system matrix is increased. As seen in this figure, the
eigenvalues appear to be advancing towards the imaginary axis in the left half of complex
splane as the system order is increased. However, it is not certain whether the
eigenvalues might cross into the right half of splane if the system order increases
greatly.
90
120 60
0.8
0.6
150 30
( )
max r3
0.4
( )
max r4
0.2
( )
max r5 180 0 0
( )
max r6
( )
max r7
210 330
240 300
270
Figure 411 Dominant root locus of the active powerfrequency control system for a microgrid as the
system order is increased from 3 through 7
architectures containing multiple inverter based DRs are studied based on eigenvalue
with the active powerfrequency and reactive powervoltage controllers for decentralized
operation. Chandorkar [60] solved the stability problem for a generalized n distributed
UPS system connected in a chain formation, where each UPS inverter was equipped with
[60] was a simplified one neglecting the frequency restoration loop of the active power
frequency controller; and furthermore the reactive powervoltage controller was absent in
that model.
The dynamic behavior is investigated in this section for the microgrid architecture
of n DRs with the infinite bus at one end and loads connected at each DR bus. It is
assumed that the tieline between any two buses i and j has an inductive reactance Xi.j.
The operation of this chain microgrid is analyzed below with gridinterfaced mode as a
special case.
The state variable schematic of the chain microgrid illustrating the DRs
connected by an inductive tieline is given in Figure 412(b). These state variables are
directly depicted for two of the units, viz., kth and k+1th in Figure 412(b), with the dashed
DR2 DRk+1
Δω k(s)
ΔVk(s) ω qk
1/Dqk bk
s
Dqtiek,k+1 ωbk
Kbβk
s
Vko
Xk,k+1 Δω hk(s)
Vk+1o
Δδk,k+1(s) 1
Xk,k+1 Pok,k+1
s
Dqtiek+1,k
Δω k+1(s)
ΔVk+1(s) ω
qk+1
1/Dqk+1 bk+1
s
ω bk+1
Kbβk+1
s
Δω hk+1(s)
For the state vector xT = [Δωh1 Δδ12 Δωh2 … Δδn1.n Δωhn ΔV1 ΔV2 … ΔVn], the system
⎡ [A11]2n1x2n1 [0]2n1xn ⎤
A = ⎢⎢ ⎥
⎥ (418)
⎣ [0]nx2n1 [A22]nxn ⎦
where the partitioned submatrices of A refer to the active and reactive power flows,
⎡ ⎤
ωb1 b1(b1/β1)ωG1Po12 0 0 … 0
⎢ 1 (b1+b2)Po12 1 b2Po23 … 0
⎥
=⎢ ⎥
0 b2(b2/β2)ωG2Po12 ωb2 b2(b2/β2)ωG2Po23 … 0
A11 (419)
0 b2Po12 1 (b2+b3)Po23 … 0
⎢ : : : ⎥
⎣ 0 0 0 … bn(bn/βn)ωGnPo.n1.n ωbn ⎦
ωq1V1o
⎡ ⎤
Dqtie12
ωq1⎛1+ D ⎞ 0 0 … 0
⎝ q1 ⎠ Dq1X12
⎢ ωq2V2o
ωq2⎛1+
Dqtie21+Dqtie23⎞ ωq2V2o ⎥
⎢ ⎥
0 … 0
Dq2X12 ⎝ D q2 ⎠ Dq2X23
A22 =
⎢ ⎥
ωq3V3o Dqtie32+Dqtie34⎞ ωq3V3o
0 ωq3⎛1+ D X … 0
Dq3X23 ⎝ D q3 ⎠ q3 34
⎢ ⎥
: : :
⎣ 0 0 0 0
⎝
Dqtie.n.n1
… ωqn⎛1+ D
qn ⎠⎦
⎞
(420)
and the eigenvalues λ11 (of A11) and λ22 (of A22) are given by the solution of
The properties of the eigenvalues of A11 and A22 are determined separately below,
Proposition 1: The sufficient conditions for all eigenvalues of Α22 of a microgrid in chain
configuration to be negative and real are ΔVk_max < Vko and ωqk, Dqk > 0 (k = 1, 2, …, n),
where ΔVk_max is the maximum droop in voltage of kth DR unit when operated in stand
Proof: Let a22.i,j represent the element in the ith row and jth column of matrix Α22, which is
For the controller constraints ωqk, Dqk > 0 (k = 1, 2, …, n), consider a diagonal
d22.i a22.i.i+1
(for i = 1, 2, …, n) (423)
d22.i+1 = a22.i+1.i
Then there exists a similar matrix B22 = D221 A22 D22 that is real symmetric given
by
⎡ ( ) ⎤
Dqtie12 V1oV2oωq1ωq2
ωq1 1+ 0 … 0
Dq1 Dq1Dq2X122
⎢ V1oV2oωq1ωq2
Dq1Dq2X122 (
ωq2 1+
Dqtie21+Dqtie23
Dq2 ) V2oV3oωq2ωq3
Dq2Dq3X232
… 0 ⎥
B22 =
⎢ 0
V2oV3oωq2ωq3
(
ωq3 1+
Dqtie32+Dqtie34
) … 0
⎥
⎢ ⎥
Dq2Dq3X232 Dq3
: : :
⎣ 0 0 …
Vn1oVnoωqn1ωqn
Dqn1DqnXn1.n2 (
ωqn 1+
Dqtie.n.n1
Dqn )⎦
(424)
It is well known that similar matrices have the same eigenvalues, and that a real
symmetric matrix has only real eigenvalues [86]. Since B22 is a real symmetric matrix
that is similar to A22, all the eigenvalues λ22 (of A22) are real.
Further, it is observed in (420) that A22 has only negative eigenvalues according
Vko Vko
if (Dqk + Dqtie.k.k1 + Dqtie.k.k+1) > X +X (425)
k1.k k.k+1
The expression in the left hand side of above inequality is a measure of total
reactive power generated by the kth DR under nominal conditions. If Qko denotes this
nominal value of generated reactive power, matrix A22 becomes diagonally dominant
Qko
if  D < Vko (428)
qk
since the ratio of Qko and Dqk gives the steadystate voltage droop in the standalone
The above inequality signifies that the condition for diagonal dominance of A22 is
when the (maximum) steadystate droop in voltage at every DR terminal under stand
alone mode is strictly less than its nominal value. This is generally true as the controller
gains are seldom designed to result in a voltage droop larger than the nominal value.
Consequently, under these conditions A22 has only negative real eigenvalues λ22.
Proposition 2: The sufficient conditions for all the eigenvalues of Α11 of a microgrid in
chain configuration to be negative and real are bk, βk, ωGk > 0 (k = 1, 2, …, n).
Proof: At first it is proved that all the eigenvalues of A11 are real and later it is proved
For bj, βj, ωGj > 0 (j = 1, 2, …, n), consider a diagonal matrix D11 given by
112
⎡ ⎤
b1(b1/β1)ωG1 0 0 0 … 0
⎢ 0
1
Po12
0 0 … 0 ⎥
D11 =
⎢ 0 0 b2(b2/β2)ωG2 0 … 0 ⎥ (430)
⎢ 0 0 0
1
… 0 ⎥
⎢ ⎥
Po23
: : :
⎣ 0 0 0 0 … bn(bn/βn)ωGn ⎦
Then there exists a similar matrix B11 = D111 A11 D11 that is real symmetric given
by
⎡ ⎤
ωb1  b1(b1/β1)ωG1Po12 0 0 … 0
As it is well known, similar matrices have same eigenvalues and a real symmetric
matrix has real eigenvalues; since B11 is a real symmetric matrix that is similar to A11, all
After having proved that all the eigenvalues of Α11 are real, it is proved that all of
transformations as
113
ωb1 b1Po12ωG1
⎡ 1 ⎤
0 0 … 0
⎢0 0 1
b2Po12ωG2 ωb2 b2Po23ωG2
0 … 0
⎥
⎢0 ⎥
… 0
A11 T= (432)
0 1 0 … 0
⎢ : : : ⎥
⎣ 0 0 0 … bnPo.n1.nωGn ωbn ⎦
1 b1Po12 0 0 … 0
⎡0 1 0 0 … 0 ⎤
where T=
⎢0 b2Po12 1 b2Po23 … 0 ⎥
(433)
⎢ 0: 0 0 1 … 0
⎥
⎣0 1 ⎦
: :
0 0 … bnPo.n1.n
Let ϕ2n1(λ11) = det(λ11 T  A11 Τ) represent the characteristic polynomial for the
λ11+ωb1 b1Po12(λ11+ωG1)
⎪ ⎪
0 0 … 0
⎪ 1 λ11 1
b2Po12(λ11+ωG2) λ11+ωb2 b2Po23(λ11+ωG2)
0 … 0
⎪
⎪ ⎪(437)
0 … 0
ϕ2n1 (λ ) =
11
0 0 1 λ11 … 0
⎪ : : : ⎪
⎪ 0 0 0 … bnPo.n1.n(λ11+ωGn) λ11+ωbn ⎪
For the above tridiagonal matrix, the determinant satisfies a recursive formula
ϕ0(λ11) = 0
⎪⎭
(438)
The eigenvalues λ11 (of A11) are given by the roots of the characteristic equation
ϕ2n1(λ11) = 0. As seen from the recursive formula in (438), the coefficients of all powers
of λ11 are positive. By the Descartes’ rule of signs [79], it is determined that the number
of positive real roots of ϕ2n1(λ11) = 0 is zero. Therefore, all the eigenvalues λ11 (of A11)
configuration are bk, βk, ωGk, ωqk, Dqk > 0 and ΔVk_max < Vko (for k = 1, 2, …, n).
Proof: It was determined in the previous two propositions that the sufficient condition for
and that for stability of reactive power control in the chain microgrid is
Combining the above two inequalities, the sufficient conditions for stability of the
voltage, angle or frequency, i.e. ΔV1 = 0, Δδg = 0 and Δωg = 0. Accordingly, if the chain
microgrid is connected to the grid utility at the nth DR, an additional state variable –
voltage angle between the nth DR and grid supply (Δδng) is considered and the state vector
now becomes xT = [Δωh1 Δδ12 Δωh2 … Δδn1.n Δωhn Δδng ΔV1 ΔV2 … ΔVn]. The modified system
⎡ ⎤
ωb1 b1(b1/β1)ωG1Po12 0 0 … 0
⎢ 1
0
(b1+b2)Po12 1 b2Po23
b2(b2/β2)ωG2Po12 ωb2 b2(b2/β2)ωG2Po23 …
… 0
0
⎥
where A11 =⎢ ⎥ , (442)
0 b2Po12 1 (b2+b3)Po23 … 0
⎢ : : : ⎥
⎣ 0 0 0 … 1 bnPong ⎦
ωq1V1o
⎡ ⎤
Dqtie12
ωq1⎛1+ D ⎞ 0 0 … 0
⎝ q1 ⎠ Dq1X12
⎢ ωq2V2o
ωq2⎛1+
Dqtie21+Dqtie23⎞ ωq2V2o ⎥
⎢ ⎥
0 … 0
Dq2X12 ⎝ D q2 ⎠ Dq2X23
A22 =
⎢ ⎥
ωq3V3o Dqtie32+Dqtie34⎞ ωq3V3o
0 ωq3⎛1+ D X … 0
Dq3X23 ⎝ D q3 ⎠ q3 34
⎢ ⎥
: : :
⎣ 0 0 0 0 … ωqn⎛1+
⎝
Dqtie.n.n1 + Dqtie.n.g
D qn ⎠⎦
⎞
(443)
As seen in the above equations, parameters Pong and Dqtie.n.g in the above matrices
correspond to the gridinterface tie line. However, it can be proved that the Proposition 1
through Proposition 3 are still valid after modifications are made in matrices A11 and A22.
116
Controllers of a DR in a Microgrid
controllers of the DRs in the microgrid is developed in this section. It is the outcome of
addition to that, dynamics of inverter internal controls that were covered in the previous
chapter as well as the IEEE P1547 standards [18] are taken into consideration for the
development of these guidelines. The design guidelines are presented below for the
controller are – b, β and ωG. Among these parameters, b can be designed to satisfy the
controlled to remain within with window of 59.3 Hz to 60.5 Hz. Therefore, the droop
accordance with their steadystate droops, (bk  βk). As it is generally desired to have the
constant equal to ½. This value would cause frequency restoration of every DR in the
microgrid to half their initial (transient) droop upon occurrence of a load change. Thus,
Kbβk = ½ (447)
This restoration gain of Kbβk = ½ implies that the value of βk is identical to gain
bk.
control is the restoration loop time constant 1/ωGk. This timeconstant is chosen taking
into consideration the inverter internal controller dynamics of the multiloop regulator. It
was observed in Chapter 3 that the inner loop bandwidth is about one 60 Hz cycle.
Therefore, the active powerfrequency controller restoration loop filter corner frequency
ωGk is chosen as 1 rad./s that is well below the inner controller’s bandwidth.
The various controller parameters that are to be designed for reactive power
voltage controller are Dq and ωq. Since the voltage droop of the DR is maximum in the
standalone mode of operation, this maximum droop should meet the IEEE P1547
118
standards governing DRs [18]. According to the IEEE P1547 standards, the permitted
range for voltages at the DR terminal is 88% to 110% (as a percentage of nominal
voltages). In per unit, the permitted range is given by 0.88 p.u. to 1.10 p.u. or in terms of
Therefore, Dqk is chosen as 10 p.u. that would give a maximum droop in ΔV is 0.10 p.u.
to 0.10 p.u..
With the control parameter Dq fixed at 10, the other parameter ωq is designed so
that the steadystate is reached in a certain duration. The timeconstant of the response of
the deviation in voltage to load change is given by the ratio 1/ωq. This timeconstant is
chosen taking into consideration the inverter internal controller dynamics of the multi
loop regulator. It was observed in Chapter 3 that the inner loop bandwidth is about one
60 Hz cycle. Therefore, the reactive powervoltage controller filter corner frequency ωqk
The load reference setpoint for an active powerfrequency controller PL_ref is set
according to the desired generation from the DR under gridinterfaced mode of operation.
It is generally assumed that the DR power rating is well above this setpoint so that the
DR is not driven into generating beyond its capacity. On the other hand, for the reactive
powervoltage controller, the load reference setpoint QL_ref is ideally initialized to zero
as this would cause the voltage (magnitude) reference to decrease below nominal voltage
for inductive loads and to increase above the nominal voltage for capacitive loads.
119
microgrid illustrating two DRs along with the grid supply is displayed in Figure 413.
The subsystems DG1+Rload1 and DG2+Rload2 are the two DRs each with its local load.
Grid Network subsystem contains the model of a stiff threephase voltage source in series
with an inductive reactance signifying the grid supply and the Thevenin impedance of the
grid network. The subsystem Tie line12 contains the modeling of the 100 yard tieline as
logic for the connection between the two DRs is generated at the block Tie12_En, and the
same for the connection between the DR1 and the Grid Network is generated at the block
Grd_En.
Figure 413 Simulink model of the microgrid showing two DRs connected to the grid supply
120
Figure 414 gives a detailed structure of the simulation model for a DR with its
local load. The DR is an averaged model of an inverter that feeds the resistive load. The
controllers described in Figure 44 and Figure 47, respectively. The system parameters
for the threephase 16 kW/208 V/60 Hz system are tabulated in Table 41. This
simulation model assumes that the bandwidth of the internal inverter control is high
operation of the generation controls. These are (i) single DR in standalone mode, (ii)
single DR in gridinterfaced mode and (iii) two DR interconnected mode. Selected results
and PLref are plotted in Figure 415 and Figure 416, respectively. The initial value of PL
ref is set at zero and therefore at a load of 0.25 p.u. the frequency is 59.9375 Hz. As seen
in Figure 415, a step change in PL from 0.25 p.u. to 0.5 p.u. gives a maximum transient
0.75
PL (p.u.)
0.5
0.25
0
5 5.25 5.5 5.75 6 6.25
60.25
60
f (Hz)
59.75
59.5
5 5.25 5.5 5.75 6 6.25
Time (s)
Figure 415 Simulation waveforms showing response to a change in load real power demand PL when the
DR is operated as a standalone unit
122
Figure 416 illustrates the response for a change in PLref from 0 p.u. to 0.5 p.u.
that reflects the current load condition. As against the response to change in PL shown in
Figure 415, the step change in PLref gives a firstorder lag response with a timeconstant
0.75
PLref (p.u.)
0.5
0.25
0
5 5.25 5.5 5.75 6 6.25
60.25
60
f (Hz)
59.75
59.5
5 5.25 5.5 5.75 6 6.25
Time (s)
Figure 416 Simulation waveforms showing response to a change in load real power ref. setpoint PLref to
match the load demand when the DR is operated as a standalone unit
voltage sources of fixed voltage amplitude and frequency in series with a tieline of
reactance Xtie = 0.1 p.u.. As it was observed that the grid frequency in the laboratory
experiment test bed was 59.96 Hz, the same was employed in simulation also. However,
the nominal frequency setpoint of the power controller is maintained at 60 Hz. The
results illustrating the response for a change in PLref from 0 p.u. to 0.5 p.u. at t = 5 s are
plotted in Figure 417. The load under this condition was 0.25 p.u.. Ideally, if the grid
123
frequency were at 60 Hz, the generated power would have followed the PLref setpoint.
However, as the frequency is 59.96 Hz, the difference between PLref and generated power
value of 0.16 p.u. for PLref = 0 p.u. and to 0.66 p.u. for PLref = 0.5 p.u.. The DR
frequency increases for a short duration to enable higher generation but eventually
1
PDR (p.u.)
0.5
0
0.5
1
5 5.25 5.5 5.75 6 6.25
60
f (Hz)
59.9
59.8
5 5.25 5.5 5.75 6 6.25
2
1.5
V (V)
1
0.5
0
5 5.25 5.5 5.75 6 6.25
Time (s)
Figure 417 Simulation waveforms showing response to a change in load power setpoint PLref from 0 p.u.
to 0.5 p.u. when the DR is connected to an infinite bus of frequency 59.96 Hz
Figure 418 illustrates the simulation waveforms when the load PL is changed
from 0.25 p.u.to 0.5 p.u.. As seen in this figure, the DR supplies only the transient change
in load but in steady state its generation returns to a value determined by PLref. Figure
419 illustrates the response for a change in PLref from 0.5 p.u. to 0 p.u. at t = 10 s. As
seen in this figure, the DR generated power returns to the earlier values when PLref is
500
400
300
200
100
0
100
200
300
400
500
9.75 9.8 9.85 9.9 9.95 10 10.05 10.1 10.15 10.2 10.25
80
60
40
20
0
20
40
60
80
9.75 9.8 9.85 9.9 9.95 10 10.05 10.1 10.15 10.2 10.25
Figure 418 Simulation waveforms illustrating the effect of load change at the DR terminals from 0.25 p.u.
to 0.5 p.u. while it is operated in gridconnected mode. Load terminal voltage is the top waveform and in
the bottom are the load current (solid), DR current (dashed) and tieline current (dashdot).
1
0.5
PDR (p.u.)
0
0.5
1
10 10.25 10.5 10.75 11 11.25
60
f (Hz)
59.9
59.8
10 10.25 10.5 10.75 11 11.25
2
1.5
V (V)
1
0.5
0
10 10.25 10.5 10.75 11 11.25
Time (s)
Figure 419 Simulation waveforms showing response to a change in load power setpoint PLref from 0.5
p.u. to 0 p.u. when the DR is connected to an infinite bus of frequency 59.96 Hz.
125
Figure 417 illustrates the results of DR1 response to a change in PL1 at its
terminals from 0.2 p.u. to 0.4 p.u.. The corresponding waveforms for DR2 that has a local
load of 0.4 p.u. are displayed in Figure 421. The load power setpoints in both units are
the same equal to 0.4 p.u.. Hence, as seen in these two figures the overall load is shared
equally by the two DRs. The final steadystate frequency is 60 Hz as the load burden of
However, the load change PL1 transient is felt by the local DR1 severely that its
generation initially increases to more than the final steadystate value. In contrast, the
farther DR2 increases gradually its generation to meet the increased load demand.
1
0.5
0
PDR1 (p.u.)
0.5
1
10 10.25 10.5 10.75 11 11.25
60
59.9
f1 (Hz)
59.8
10 10.25 10.5 10.75 11
2
1.5
1
V1 (V)
0.5
0
10 10.25 10.5 10.75 11 11.25
Time (s)
Figure 420 Simulation waveforms showing response of DR1 to a change in load PL1 from 0.2 p.u. to 0.4
p.u. when the two DRs are interconnected.
Figure 422 illustrates the DR1 load terminal voltage, DR1 current and the tieline
current between the two DRs. As seen in the figure, since the two interconnected DRs are
supplying their local loads, the current through the intertie is zero.
126
PDR2 (p.u.)
0.5
0
0.5
1
10 10.25 10.5 10.75 11 11.25
f2 (Hz) 60
59.9
59.8
10 10.25 10.5 10.75 11 11.25
2
V2 (V)
0
10 10.25 10.5 10.75 11 11.25
Time (s)
Figure 421 Simulation waveforms showing response of DR2 to a change in load PL1 from 0.2 p.u. to 0.4
p.u. when the two DRs are interconnected.
500
400
300
200
100
0
100
200
300
400
500
10.5 10.55 10.6 10.65 10.7 10.75 10.8 10.85 10.9 10.95 11
80
60
40
20
0
20
40
60
80
10.5 10.55 10.6 10.65 10.7 10.75 10.8 10.85 10.9 10.95 11
Figure 422 Simulation waveforms illustrating the two interconnected DRs supplying their local loads with
zero power flowing along the tieline. Load terminal voltage is the top waveform and in the bottom are the
DR current (solid) and tieline current (dashdot).
127
4.6 Summary
frequency and reactive powervoltage controllers for the inverter based DRs. These are
communication between the DRs. Smallsignal models were developed for microgrids
signal) behavior of an n DR chain microgrid is carried out using eigenvalue analysis and
eigenvalue investigations considered both the real and reactive power flows in the
was also covered. It was proved by means of mathematical propositions that the n DR
chain microgrid is stable for all positive values of generation controller gains with the
maximum droop limited to be less than the nominal value. Guidelines were provided for
design of the active powerfrequency and reactive powervoltage controllers. Besides the
results of dynamic behavior analysis, the IEEE P1547 performance specifications [18]
were also utilized in the design process. Finally, selected results are presented from
The dynamic behavior of a chain microgrid had been investigated in the previous
microgrid, distribution networks of other architectures are also widespread. This chapter
investigated on the operation of the microgrid and preliminary results are presented. As
in the previous chapter, the following assumptions are made in the analysis of dynamic
behavior —
(ii) Inverter based DRs operate within their maximum capacity limits
(iii) Dynamics of the inverter internal controls are fast as compared to the
(iv) Tielines between any two sources in the microgrid are purely inductive
in nature
Parallel
The previous chapter covered the sufficient conditions for a microgrid having a
chain structure have been derived. This section covers the dynamic behavior for the
in the figure, the kth DR is connected to the PCC by an inductive tieline of reactance Xk.
PCC (Vp,δp)
X1 X2 Xn
Figure 51 Singleline diagram of a microgrid parallel structure consisting of several DRs
Proposition 4: (i) The incremental change in voltage magnitude ΔVp at the PCC in a
(ii) The incremental change in phase angle Δδp at the PCC in a parallel microgrid is a
(iii) Also, the incremental change in frequency Δωp at the PCC in a parallel microgrid is
Proof: (i) As seen in Figure 51, n DRs are connected to the infinite bus through a star
point, known as point of common coupling (PCC). The voltage at the PCC is dependent
on the voltages at all the sources connected to it by a weighted average. This PCC
~
⎛Xprll⎞ ~ ⎛Xprll⎞ ~ ⎛Xprll⎞ ~
Vp = ⎜ X ⎟ V1 + ⎜ X ⎟ V2 + … + ⎜ X ⎟ Vn , (51)
⎝ 1⎠ ⎝ 2⎠ ⎝ n⎠
In the above equation, the expression for equivalent parallel inductance Xprll is
obtained from
1 1 1 1
Xprll = X1 + X2 + … + Xn (52)
Equation (51) for phasor voltage PCC can be rewritten in terms of complex
exponential quantities as
Expansion of complex exponential functions into real and imaginary parts gives
131
⎛Xprll⎞ ⎛Xprll⎞
Vp (cos(δp) + j sin(δp)) = ⎜ X ⎟ V1 (cos(δ1) + j sin(δ1)) + ⎜ X ⎟ V2 (cos(δ2) + j sin(δ2))
⎝ 1⎠ ⎝ 2⎠
⎛Xprll⎞
+ … + ⎜ X ⎟ Vn (cos(δn) + j sin(δn)) (54)
⎝ n⎠
Practically the difference between the phase angles of any two voltage sources
connected by an inductive tieline is a very small value; and a phase angle reference can
cos(δk) = 1 ⎫
⎬ ∀ k ∈ 1, 2, …, n (55)
sin(δk) = δk ⎭
As a particular case, if the infinite bus is one of the sources connected to the PCC,
it can be chosen as the phase angle reference and therefore its phase angle is zero.
The expression for voltage phasor at the PCC would then become
(56)
Equating the real parts on left and righthand sides of (56), the magnitude
condition is determined as
Vp = c1 V1 + c2 V2 + … + cn Vn (57)
⎛Xprll⎞
where ck = ⎜ X ⎟ ≥ 0 and ∑ck = 1.
⎝ k⎠
n
ΔVp =
∑ ⎛∂Vp⎞
⎜ ⎟ ΔVk
⎝∂Vk⎠
k=1
⎛Xprll⎞
where ck = ⎜ X ⎟ ≥ 0 and ∑ck = 1. This is valid under the assumption that the incremental
⎝ k⎠
(ii) Equating the imaginary parts on left and righthand sides of (56), the phase angle
condition is determined as
⎛Vk⎞
where dk' = ⎜V ⎟ ck ≥ 0 and ∑dk' = 1.
⎝ p⎠
If incremental changes are allowed in phase angles of voltages, since δk << 1 (for
k = 1, 2, …, n), the relationship between the incremental changes in phase angles is given
by
n
Δδp =
∑ ⎛⎛∂δp⎞ ⎛ ∂δp ⎞
⎜⎜ ⎟ Δδk + ⎜
⎝⎝∂δk⎠ ⎝∂dk'⎠
⎞
⎟ Δdk'⎟
⎠
k=1
⎛Vko⎞
where dk = ⎜V ⎟ ck ≥ 0 and ∑dk = 1 . This is valid under the assumption
⎝ po⎠
(iii) Since the coefficients ck and dk (for k = 1, 2, …, n) are independent of time, the
expression for incremental change in frequency Δωp at the PCC, after differentiating with
respect to time on both sides of the phase angle relationship given above, is determined
as
d
Δωp = dt (Δδp) = d1 Δω1 + d2 Δω2 + … + dn Δωn (511)
of the incremental changes in frequencies of the sources connected to it. This weighting
factor is determined from the tieline inductance and voltage magnitudes on either side of
the tieline.
With the help of the Proposition 4 stated above, it is now possible to study the
system matrix of the parallel microgrid. The state variable schematic of the parallel
reactance Xk is given in Figure 52. The state vector x can be chosen such that its
transpose is given by xT = [Δωh1 Δδ1p Δωh2 Δδ2p Δωh3 … Δδn1.p Δωhn ΔV1 ΔV2 … ΔVn].
The system matrix, A, from which the stability of the interconnected parallel
microgrid can be determined. The matrix A for the microgrid in Figure 52 is given by
⎡ [A11]2n1x2n1 [0]2n1xn ⎤
A = ⎢⎢ ⎥
⎥ (512)
⎣ [0]nx2n1 [A22]nxn ⎦
where A11 and A22 represent the active and reactive power flows, respectively. Their
eigenvalues, λ11 (of A11) and λ22 (of A22), can be determined independently.
ΔVk(s) ω qk Δω k(s)
1/Dqk bk
s
Dqtiek,p ω bk
Kbβk
s
Δω hk(s)
ck
Δδ k,p(s) 1
Pok,p
s
dk
ΔVp(s)
Vko
Δω p(s)
Xk,p
Figure 52 Smallsignal statevariable schematic under zero input conditions of a microgrid having DRs in
parallel
Moreover, the sum of real powers flowing from all DRs to the PCC is equal to
zero.
The partitioned submatrices of A refer to the active and reactive power flows,
⎢ 0 1 1 1 n n o1p d2 … (bn1dn1bndn)Po.n1.p dn
⎥
⎢ d (b d b d )P ⎥
0 ωb2 … 0 0
A11 = (516)
1 1 1 n n o1p (1d2) … (bn1dn1bndn)Po.n1.p dn
⎢ . . . … . . ⎥
⎣ 0 b (b /β )ω P n n n Gn o1p 0 … bn(bn/βn)ωGnPo.n1.p ωbn ⎦
135
⎣ c1ωqnVno
DqnXn
c2ωqnVno
DqnXn
Dqtie.n.p cnωqnVno
… ωqn⎛1+ D ⎞+ D X
⎝ qn ⎠ qn n
⎦
(517)
The properties of the eigenvalues of A11 and A22 are determined separately below,
Proposition 5: The sufficient conditions for all eigenvalues of Α22 of a parallel microgrid
to have negative real parts are ΔVk_max < Vko and ωqk, Dqk > 0 (k = 1, 2, …, n), where
ΔVk_max is the maximum incremental change in voltage of kth DR unit when operated in
Proof: By applying Gerschgorin theorem [86] on Α22 = (a22.i,j), it is well known that the
n
eigenvalues of A22 lie in disks with centers a22.i.i and radii ρ22.i = ∑ a22.i.j (where i = 1, 2,
j=1
…, n, and j ≠ i). With all of its diagonal elements strictly negative, A22 has only
eigenvalues with negative real parts if it is diagonally dominant under the following
ωqiVio n ω V
⎛ Dqtie.i.p⎞ qi io
i.e., ωqi⎜1+ D ⎟ > D X
⎝ qi ⎠ qi i
∑ cj = DqiXi (519)
j=1
136
Vio
i.e., (Dqi + Dqtie.i.p) > X (520)
i
⎛Vio  Vpo⎞
i.e., Vio ⎜ X ⎟ > Dqi Vio (522)
⎝ i ⎠
The expression in the left hand side of above inequality is a measure of total
reactive power generated by the ith DR under nominal conditions. If Qio denotes this
nominal value of generated reactive power, matrix A22 becomes diagonally dominant
Qio
i.e.,  D < Vio (523)
qi
The above inequality signifies that the condition that diagonal dominance of A22
is when the (maximum) standalone droop in voltage at every DR terminal is strictly less
than its nominal value. This is generally true as practically the controller gains are never
designed to result in a voltage droop larger than the nominal value. Consequently, under
Proposition 6: The sufficient conditions for all eigenvalues of A11 of a parallel microgrid
ωbk + bkPokp = C1 ⎪⎫
⎬ (for k = 1, 2, …, n)
bkωGkPokp = C2 ⎪⎭
Proof: The dense matrix A11 is characteristic of the parallel microgrid where every DR
137
unit has some form of coupling to all remaining units. This coupling obscures and makes
it difficult to evaluate the properties of matrix A11. However, the matrix A11 can be made
(ωb1+b1Po1p) b1ωG1Po1p
⎡ ⎤
0 0 … b1Po1p
⎢ 1
0
0
0
0
(ωb2+b2Po2p) b2ωG2Po2p …
0 … 1
b2Po2p
⎥
B11 =⎢ ⎥ (525)
0 0 1 0 … 1
⎢ . . . . … . ⎥
⎣ b11.2n1.1 b11.2n1.2 b11.2n1.3 b11.2n1.4 … b11.2n1.2n1 ⎦
and the transformation matrix,
⎡0 ⎤
1 b1Po1p 0 0 … 0
⎢0 1
0
0
1
0
b2Po2p
… 0
… 0
⎥
=⎢
… 0 ⎥
P11 (526)
0 0 0 1
⎢. . . . … . ⎥
⎣ d 1 (dnbnd1b1)Po1p d2 (dnbnd2b2)Po2p … d ⎦ n
In the matrix B11 = (b11.i,j), the elements of the 2n1th row are determined as
⎛ n ⎞
and b11.2n1.2n1 =  ⎜ωbn + bnPonp  ∑ dhbhPohp⎟ (528)
⎜ ⎟
⎝ h=1 ⎠
As seen in (525), all rows/columns except one in B11 = (b11.i,j) already have a
  0 0 … +
⎡ ⎤
⎢ +
0
0
0
0

0

…
…

+
⎥
sign(B ) = ⎢
11
⎥ (529)
0 0 + 0 … 
⎢ . . . . … . ⎥
⎣ x x x x … x ⎦
The entries labeled ‘x’ in sign(B11) are not known a priori. Nevertheless, once
these entries are known the sign pattern of matrix B11 is adequate to demonstrate
qualitative stability known as sign stability. A matrix B11 is called sign stable if each
matrix C11 of the same qualitative (or sign) pattern as B11 (sign b11.i,j = sign c11.i,j for all
i,j) is stable regardless of the magnitudes of b11.i,j [87]. The stability of A11 can be judged
from that of B11 as the eigenvalues of similar matrices A11 and B11 are identical.
According to Jeffries et al [87, Theorem 2], the 2n1 x 2n1 real matrix B11 =
(b11.i,j) is sign stable if and only if it satisfies the following five conditions:
The offdiagonal entries b11.ij and b11.ji must not be of same sign.
elements V = {1, 2, 3, …, 2n1} and the edge set comprises the nonzero
(iv) In every RB11coloring of the undirected graph GB11, all vertices are
black.
elements V = {1, 2, 3, …, 2n1}, the edge set comprises the ordered pairs
and b11.i,j ≠ 0 ≠ b11.j,i} and rows having nonzero diagonal entries form the
set RB11 = {i: b11.i,i ≠ 0}. According to the definition [87], elements of RB11
are painted black and no black vertex is allowed to have precisely one
white neighbor, and each white vertex must have atleast one white
zero.
It is observed that the unique solution that satisfies all the above five conditions,
(i) through (v), is the last row offdiagonal entries of B11 being equal to zero; i.e. b11.2n1,j
= 0 (for j = 1, 2, …, 2n2).
In this case,
140
and therefore, condition (i) is satisfied for the last row element also as shown by
⎛ n ⎞ n
⎜ ⎟
b11.2n1.2n1 =  ωbn + bnPonp  ∑ dhbhPohp =  ∑ ωbk (531)
⎜ ⎟
⎝ h=1 ⎠ h=1
Condition (ii) is verified by inspection. For condition (iii), the directed graph DB11
is illustrated in Figure 53. It has V = {1, 2, 3, …, 2n1} as its vertex set and ED = {(1,2),
(1,2n1), (2,1), (2,2n1), (3,4), (3,2n1), (4,3), (4,2n1), (5,6), (5,2n1), (6,5), (6,2n1), …,
(2n3,2n2), (2n3,2n1), (2n2,2n3), (2n2,2n1)} as its edge set. As seen in the figure,
dark lines represent the elements of edge set. Light grey lines are also displayed in Figure
53 that represent ordered pairs involving the last (i.e. 2n1th) row of B11. These light grey
lines are not part of the edge set ED as presence of even a single ordered pair involving
the last (i.e. 2n1th) row of B11 in the edge set would bring about a 3cycle. Thus, the only
2n2).
141
2n1
b11.1,2n1 b11.2n2,2n1
b11.2n1,4 b11.2n1,2n2
b11.2n1,3
b11.2n1,2 b11.2n1,2n3
For condition (iv), Figure 54 shows the undirected graph GB11 that has V = {1, 2,
3, …, 2n1} as its vertex set, EG = {(1,2), (3,4), (5,6), …, (2n3,2n2)} as its edge set and
RB11 = {1, 3, 5, …, 2n1}. As seen in Figure 54, the edges in GB11 correspond to the 2
cycles in GB11. Besides, the elements of RB11 are painted black. However, as no black
vertex is allowed to have precisely one white neighbor, and each white vertex must have
atleast one white neighbor according to the definition [87], vertices {2, 4, 6, …, 2n2} are
also painted black to prevent the elements of RB11 from having a single white neighbor.
As a consequence, in every RB11coloring of the undirected graph GB11, all vertices are
Likewise, condition (v) can be proved for nonnegative control parameters as the
[87] since the determinant of B11 has some nonzero term in its expansion and condition
142
(iii) is satisfied.
2n1
1 2 3 4 2n3 2n2
Thus, it is proved that the sufficient conditions for B11 (or A11) to be a stable
ωbk + bkPokp = C1 ⎫⎪
or ⎬ (for k = 1, 2, …, n) (533)
bkωGkPokp = C2 ⎭⎪
flow in the parallel microgrid are more rigid as compared to those of a chain. These are
very conservative on stability and are important as they have a deeper physical meaning.
the PCC of a parallel microgrid is a weighted average of the excursions of the same
143
quantities in all constituent DRs. Hence, the nearest DR to the PCC is the dominant DR
the sufficient conditions in (533) recommend the following criteria in the design of a
parallel microgrid —
their droop characteristic (that has a slope of b  β), the DR that has a
configuration are ωqk, Dqk > 0, ΔVk_max < Vko (for k = 1, 2, …, n) and
ωbk + bkPokp = C1 ⎪⎫
⎬ (for k = 1, 2, …, n)
bkωGkPokp = C2 ⎭⎪
Proof: It was determined in the previous two propositions that the sufficient conditions
ωbk + bkPokp = C1 ⎪⎫
⎬ (for k = 1, 2, …, n)
bkωGkPokp = C2 ⎪⎭
Therefore, the sufficient conditions for stability of the overall parallel microgrid
Operation
voltage, angle or frequency, i.e. ΔV1 = 0, Δδg = 0 and Δωg = 0. If the parallel microgrid
is connected from the PCC to the infinite bus of voltage E through a tieline of
⎫⎪
n
∑ck = 1  c g
⎬
k=1
(534)
⎪⎭
n
∑dk = 1  dg
k=1
where ck and dk are computed after taking into consideration parameters (Vg and Xg) of
and PCC (Δδnp) is considered as an additional state variable and thus the state vector now
becomes xT = [Δωh1 Δδ1p Δωh2 Δδ2p Δωh3 … Δδn1.p Δωhn Δδn.p ΔV1 ΔV2 … ΔVn]. The system matrix
is then determined as
⎡ (1d ) ⎤
ωb1 b1(b1/β1)ωG1Po1p 0 … 0 0
⎢ 0 1 b1(1d1)Po1p d2 … dn bndnPo.n.p
⎥
⎢ d ⎥
0 ωb2 … 0 0
where A11 = , (536)
1 b1d1Po1p (1d2) … dn bndnPo.n.p
⎢ . ⎥
⎣ d ⎦
. . … . .
1 b1d1Po1p d2 … (1dn) bn(1dn)Po.n.p
⎣ c1ωqnVno
DqnXn
c2ωqnVno
DqnXn
Dqtie.n.p cnωqnVno
… ωqn⎛1+ D ⎞+ D X
⎝ qn ⎠ qn n
⎦
(537)
ωqiVio n ω V
⎛ Dqtie.i.p⎞ qi io
ωqi⎜1+ D ⎟ > D X
⎝ qi ⎠ qi i
∑ cj = DqiXi (1  cg) (for i = 1, 2, …, n) (538)
j=1
Likewise, Proposition 6 can be proved for this case after making the similar
⎡0 ⎤
1 b1Po1p 0 0 … 0
⎢0 1
0
0
1
0
b2Po2p …
… 0
0
⎥
P11 =⎢ ⎥ (539)
. . . . … .
⎢ d d1b1Po1p d2 d2b2Po2p … dnbnPo.n.p ⎥
⎣0 ⎦
1
d1 0 d2 … dn
146
In the case of reactive power flow in microgrid, the sufficient conditions for
stability of parallel structure are identical to those of the chain architecture dealt in the
previous chapter. The sufficient conditions derived for stability of real power flow in the
parallel microgrid (533), on the other hand, are rigid as compared to those of a chain. It
is to be noted that for a parallel microgrid, it is not necessary to satisfy the conditions
ωbk + bkPokp = C1 ⎫⎪
⎬ (for k = 1, 2, …, n) , (533)
bkωGkPokp = C2 ⎭⎪
to prevent instability. This is despite the fact that under such an event, one or more of
conditions (i) through (v) among the necessary and sufficient conditions for sign stability
parallel microgrid. Consider 5 DRs to be operating in the baseline case with identical
parameters as in Table 41 that are designed in accordance with the guidelines provided
in the previous chapter. The tielines between each unit and the PCC are also assumed to
be identical for all DRs (equal to 0.1 p.u. reactance) so as to guaranty that conditions in
(533) are met by this baseline case. The eigenvalues of the system under this situation
are determined from MathCAD® software package as in Table 51. As seen in this table,
the baseline case is stable as it has all eigenvalues in the left half of the splane. The most
dominant eigenvalue is at 0.968. However, it is also observed from the eigenvalues that
there are repeated values among them. The eigenvalues of a chain microgrid, in contrast,
147
or more of conditions (i) through (v) among the necessary and sufficient conditions for
sign stability [87, Theorem 2]. Jeffries et al have also proposed Lemmas [87, Lemma 1 –
Lemma 3] that suggest possibilities of migration of eigenvalues into the right half of s
plane when the conditions are violated. It is observed that conditions (ii) and (iii) can get
easily violated.
The parameter chosen as the variable for this analysis is ωG5. By increasing or
decreasing ωG5 from 1 rad./s, these conditions (ii) and (iii) get violated. Using
is varied. It is observed that the dominant eigenvalue of 0.968 stays the same regardless
of the increase in ωG5. On the other hand, if ωG5 is decreased below 1 rad./s, the
migration of the dominant eigenvalue that was initially at 0.968 is towards the imaginary
axis of the complex splane as illustrated in Figure 55. However, it is verified that this
eigenvalue does not reach the imaginary axis regardless of the decrease in the value of
parameter ωG5.
148
90
120 60
0.08
0.06
150 30
0.04
0.02
210 330
240 300
270
Figure 55 Locus of the dominant eigenvalue of the real power flow in a 5DR parallel microgrid as ωG5 is
reduced below 1 rad./s.
This section has thus proved for a parallel microgrid of 5 DRs that the sufficient
conditions in (533) of Proposition 6 even if violated do not cause any instability in the
microgrid. However, it remains to be proven that the violation does not cause any
Behavior of Microgrid
were carried out on the dynamic behavior of the microgrid for the chain and parallel
topologies. In this section, the dynamic behavior is investigated for the simplest of the
mesh networks, viz., a 3DR mesh. It consists of 3DRs with interconnection through an
inductive tieline Xij between units i and j. Consider a 3unit mesh network as shown in
149
Figure 56(a). Based on Proposition 4, the following Corollary 4.1 is proved on its small
⎛X1 X2 + X2 X3 + X3 X1⎞
X12 = ⎜ ⎟ ,
⎝ X3 ⎠
⎛X1 X2 + X2 X3 + X3 X1⎞
X23 = ⎜ ⎟
⎝ X1 ⎠
⎛X1 X2 + X2 X3 + X3 X1⎞
and X31 = ⎜ ⎟
⎝ X2 ⎠
Proof: The smallsignal real power generated by the kth DR unit of the 3unit mesh
3
ΔPk = ∑( Po.k.i (Δδk  Δδi)) for k = 1, 2, 3 (540)
i=1
150
(ΔP1, ΔQ1)
(ΔV1, Δδ1)
X12 X31
X1 X2 X3
(ΔV1, Δδ1) (ΔV2, Δδ2) (ΔV3, Δδ3)
(b)
On the other hand, the smallsignal real power generated by the kth DR unit of the
Substituting the expression for Δδp from Proposition 4 with n = 3, the above
equation is simplified to
3
ΔPk = Po.k.p ∑(di (Δδk  Δδi)) for k = 1, 2, 3 (542)
i=1
Substituting the values of base powers Po.k.i, Po.k.p and coefficients di in the above
equation,
⎛ X1 X2 X3 ⎞
Xprll = ⎜X X + X X + X X ⎟ (545)
⎝ 1 2 2 3 3 1⎠
Consequently,
Xk Xi
Xki = X for k = 1, 2, 3; i ≠ k (546)
prll
⎛X1 X2 + X2 X3 + X3 X1⎞
i.e. X12 = ⎜ ⎟ , (547)
⎝ X3 ⎠
⎛X1 X2 + X2 X3 + X3 X1⎞
X23 = ⎜ ⎟ (548)
⎝ X1 ⎠
⎛X1 X2 + X2 X3 + X3 X1⎞
and X31 = ⎜ ⎟ (549)
⎝ X2 ⎠
Thus, the meshparallel analogy of a 3unit microgrid has been proved in the case
of real power flow. Likewise, the same can also be derived for reactive power flow, but is
not covered here for purpose of brevity. The above relation between reactances in
Corollary 4.1 is identical to the well known stardelta transformation of network theory
[88].
152
The dynamic behavior of microgrid was analyzed assuming purely inductive tie
lines. This is in accordance with the traditional power flow studies that focus on high
voltage transmission lines with R/X ratio small enough that it can be neglected [89].
buildings. In such installations, the R/X ratio cannot be neglected as indicated by the
tables in Figure 57 [90]. In this section, the effect of variation in R/X ratio on
eigenvalues of the microgrid is analyzed for the simple case of a single DR connected to
Figure 58(a) displays the schematic of a single DR connected to the infinite bus
through a tieline of inductance X1,g and R/X ratio that is denoted by σ. The smallsignal
statevariable schematic of this system under zero input conditions is given in Figure
58(b). As seen in the figure, a finite value of σ creates coupling between the real and
‡
Source: [90]
154
PL1, QL1
σX1,g X1,g
Infinite Bus
V1∠δ1 E∠0o
DR1
(a) Singleline diagram
ΔV1(s) Δω 1(s)
ωq1
1/Dq1 b1
s
Dqtie1,g ω b1
1 + σ2 Kbβ1
s
σDqtie1,g Δωh1(s)
1+ σ 2 Po1,g
1 + σ2 Δδ1(s) 1
σPo1,g s
1 + σ2
(b) Smallsignal statevariable schematic under zero input conditions
Figure 58 Single DR connected to the infinite bus through a tieline of R/X ratio σ
d b1 b1/β1 ωG1
dt Δωh1 =  ωb1 Δωh1  1 + σ2 (Po1g Δδ1g + σDqtie1g ΔV1) , (550)
d b1
dt Δδ1g =  Δωh1  1 + σ2 (Po1g Δδ1g + σDqtie1g ΔV1) (551)
d ⎛⎛ 1 ⎛Dqtie1g⎞⎞ σ ⎛Po1g⎞ ⎞
and ΔV1 =  ωq1 ⎜⎜1 + 2⎜ ⎟ ⎟ ΔV1 2⎜ ⎟ Δδ1g⎟ (552)
dt ⎝⎝ 1 + σ ⎝ Dq1 ⎠⎠ 1 + σ ⎝ Dq1 ⎠ ⎠
b1(b1/β1)ωG1σDqtie.1.g
⎡ ⎤
b1(b1/β1)ωG1Po1g
ωb1
1+σ2 1+σ2
A=
⎢ 1 b1Po1g b1σDqtie.1.g ⎥ (553)
⎢ 1+σ2 1+σ2
⎥
⎣0 ωq1σPo1g
Dq1(1+σ2)
ωq1 ⎛1 +
⎝
1 ⎛Dqtie1g
1+σ2⎝ Dq1 ⎠⎠ ⎦
⎞⎞
155
As observed in Figure 58(b), the coupling between real and reactive power loops
(for σ ≠ 0) makes it no longer possible to decouple matrix A into A11 and A22
corresponding to the real and reactive power controls. Therefore, the eigenvalues of the
behavior for σ ≠ 0.
To investigate the effect of variation in R/X ratio on the dynamic behavior of the
single DR connected to infinite bus system, the dominant eigenvalues of A are plotted in
Figure 59 for typical system parameters as σ varies from 0 through 5 using MathCAD®
software package. As seen in Figure 59, the three eigenvalues advance towards the
imaginary axis in the left half of complex splane as the system order is increased. Thus,
the dynamic behavior of the single DR connected to infinite bus is greatly affected by
90
120 10 60
8
150 6 30
4
EIGi , 0 2
EIGi , 1 180 0 0
EIGi , 2
210 330
240 300
270
Figure 59 Dominant eigenvalues of the single DR connected to infinite bus as R/X ratio is increased from
0 through 5.
156
5.4 Summary
signal models were developed for the parallel connection of DRs in a microgrid to a
point of common coupling (PCC). It was proved that the excursion in voltage/phase
angle/frequency at the PCC are a weighted average of the excursions of the same
quantities at the DRs that are connected to it. Further, preliminary results for stability of
the parallel microgrid were reported in the form of sufficient conditions. The qualitative
determining the excursions in voltage, phase angle and frequency at the PCC. In such a
scenario, the sufficient conditions for stability of a parallel microgrid recommend the
their droop characteristic (that has a slope of b  β), the DR that has a
violation of sufficient conditions does not endanger the microgrid. However, it remains to
157
be proven that n DR parallel microgrids are stable over a large range of controller gains.
The mesh topology of microgrids was analyzed for the simplest case of a 3DR mesh
network based on its smallsignal analogy to the 3DR parallel network. Finally, since a
nonzero R/X ratio is typical of wires in medium voltage distribution systems, effects of
R/X ratio were investigated in the active and reactive power flow. The microgrid system
models developed in this chapter can also be applied elsewhere to investigate the impact
systems.
158
Microgrid
The preceding chapters dealt with the control strategy of the DR. These included
regulators for the load bus voltage as well as for the generated power of the DR. While
the voltage controller regulates the bus voltage against load variation as well as
imbalances due to unequal loads and voltage sags, the external generation controller
regulates the power flowing from the DR into the microgrid. The design of these
controllers and proof of the concept has been presented in the form of large signal
in the earlier chapters of this thesis, and serve as a testbed platform to verify other control
strategies on hardware [66,91], a laboratory microgrid has been constructed [10,73]. This
chapter deals with the experimental results obtained on the laboratory microgrid. The
verify the controller described in this proposal, is displayed in Figure 61. This is same as
the schematic shown in Figure 13, but with protection apparatus necessary for safety of
equipment and personnel included. As seen in Figure 61, the microgrid consists of two
threephase pulsewidth modulated (PWM) inverter based DRs (viz., DG1 and DG2)
each having a generation capacity of 22 kVA and rated for threephase/480 V/60 Hz.
Each inverter is equipped with LCL filters to attenuate the PWM switching ripples. The
interconnections between various DRs in the microgrid is made at 208 V level, and this is
facilitated by means of ΔY transformers that step down 480 V generated by each of the
DRs to the 208 V level. The two DRs are interconnected by a 100 yard tieline with load
location, Load12. The grid is connected to this network at the busbar B1, which is also
the load terminal of DG1 (Load1). Further, circuit breakers (CB) are provided in each
line for protection purposes, and contactors (CN) are included along those lines that
would need reconnection when both ends are active. A static switch SS1 consisting of
backtoback thyristors is employed for connecting the microgrid to the grid mains
network as it would enable seamless transfer of the operation of the microgrid between
75 yard
cable
CN2
CB4
CB5*
Load12
B Busbar
BS Bypass Switch B2 CB6*
C Capacitor Tieline
CB Circuit Breaker w/ Shunt Trip Load2
Interconnect
CB* Main Circuit Breaker and
Breakers for Individual Loads 25 yard
T3
Lt2 Lf2
CN Contactor cable
FD Fused Disconnect
L Inductor YΔ Cf2 DG2
SS Static Switch B3
T Transformer
R Resistor
Figure 61 One line diagram of the laboratory microgrid consisting of two DRs
160
161
(DG1/DG2 in Figure 61) along with its LC filter network is illustrated in Figure 62. The
following equipment has been employed in the construction of the lab microgrid.
each DR is 15 kW of real power and an equal amount to reactive power. For this reason,
the motordrive inverter is overrated for 37 kW (i.e., 50 HP) to withstand possible over
voltages and overcurrents that occur in the event of fault conditions. A common dcbus
configuration model of AllenBradley drive is chosen for this purpose. The nameplate
details of the motor drive inverter [92] are tabulated in Table 61. Dc input voltage to the
inverter is obtained from two programmable dc power supplies connected in series, each
Current 63 A
Voltage 0 – 575 V
Current 57.2 A
Filter Selection:
Each of the inverter based DRs has an LC filter that is designed to attenuate the
switching ripples of the PWM inverter. The values of Lf and Cf are chosen such that the
overall filter size is optimized for a particular corner frequency. The designed values of
filter parameters for the DRs of the laboratory microgrid are tabulated in Table 62.
Rf 0.21 Ω
Cf 30 μF
fc 933 Hz
‡
Source: [92]
INV1/2
VDC+
5 mH/ 0.97 mH/
phase phase
DC1_1/DC2_1
Lt1/2 Lf1/2
A
B
C
VDC
Cf1/2
30 mF/phase
VC1/2 , E1/2
CTRL1/2
IL1/2 , IT1/2
DSP1/2
C Capacitor
CTRL Control and Interface Board
DC PQ50020 DCPower Supply
DSP Motorola56F805 DSPEVM
INV AB Drive Inverter
L Inductor
Figure 62 Circuit schematic of the PWM inverter based DR unit with its LCL filter
163
164
Transformer Selection:
A ΔY transformer with each DR inverter would allow connection of both single
and threephase loads to the DR. The specifications of this transformer in the setup are
make  Cat. #QO124L100G) that has 24 poles for various threewire and fourwire loads.
Furthermore, a main circuit breaker with shunt trip (SquareD make  Cat.
#KAL3G1001021) is provided before the load center for disconnecting the power supply
in the event of large unbalance (>10%) in the terminal voltage, phaseloss, overvoltage
or undervoltage.
Figure 63 illustrates the threephase schematic of the wiring layout at the load
terminal busbar. As seen in the figure, the 480 V (Δ) primary side of the transformer
consists of the DR and its LCL filter network and its 208 V (120 V  Y) secondary side of
the transformer is connected to the fourconductor busbar. The Yside of the threephase
Furthermore, Figure 63 displays the schematic of the busbar to the load and tie
line. The various singlephase and threephase loads are connected through a load center,
which consists of a main circuit breaker (MCB) and individual cicuit breakers (LCBs) for
each load. It is to be noted that a conductor of larger size is employed for the neutral
wiring as compared to the three phases. This is due to the possible overloading of the
neutral phase in the event of several singlephase or harmonic loads are present in the
network to the grid mains (refer Figure 61 for a oneline diagram of the microgrid). The
static switch consists of backtoback thyristors in each phase [91] with associated
snubber circuitry. A synchronization logic and thyristor firing circuitry provides the gate
signals to the static switch on the basis of the threephase voltage waveform information
X1
B3 H1
A”
AWG #3 H2
B” E2
Tieline
C”
X2 X0 X3
AWG #2 H3
n c b a 45 kVA / Δ Y
MCB6 Transformer #3
LCB62
B Busbar
MCB Main CircuitBreaker
LCB61
LC Load Center
LCB Breakers for Individual Loads
LC2
Figure 63 Wiring layout of the transformer, loads and the tieline connections to the busbar (to be modified for contactor)
166
167
The PWM switching logic of each VSI based DR is generated using a digital
signal processor (DSP) [72]. Some important features of the DSP are tabulated in Table
64.
are programmed in the DSP. Motorola Codewarrior™ and PCMaster™ are used for
Figure 64 Distributed generator DG1 supplying power to its local load as well as tied to the microgrid
169
Figure 65 Connections of the distributed generators DG2 and DG3 ‡ to the microgrid
‡
DG3 is not operational and hence is not shown in the circuit schematic of Figure 61
170
Figure 67 Static switch interface for utility/grid interconnection of the microgrid ‡
‡
H. Zhang [91]
172
Several experiments were carried out in the laboratory microgrid for testing the
inverter internal controls and the external generation controls that were developed in the
previous chapters. Few selected waveforms are presented in this section demonstrating
the multiloop regulator are presented. The controllers were implemented in the
Motorola® 56F805 digital signal processor (DSP) evaluation board [72] whose features
Figure 68 illustrates the performance of the voltage regulator for an incremental
step change in voltage command from 70% to 100% of the rated voltage of 480 VRMS.
Waveforms displayed in Figure 68 are the αβ components of reference voltage, actual
voltage that tracks the reference and the error between the reference and actual values.
These are internal variables in 56F805 DSP and have been obtained from PC Master™ of
reaches close to the steady state in about three 60 Hz cycles. During this transient, the
actual voltage of filter capacitor vCf(t) across phases A and C was captured using a scope
and is displayed in Figure 69. This was followed by tests under balanced threephase
in p.u.
in p.u.
Time (ms)
Figure 68 Operation of the voltage control loop for vCf with the response vCf for a complex exponential
step input voltage reference vCf†. DSP obtained from PC Master™ of Metrowerks® CodeWarrior™.
Figure 69 Experimental waveform of line voltage vCf,AC illustrating the response for a step change in
voltage reference vCf† from 70% to 100% of rated voltage of 480 VRMS
The test in this case involves a threephase 10 kW resistive load. Figure 610
illustrates the steadystate voltage waveforms at the load terminals across phases ac and
174
Figure 610 Load terminal lineline voltage ac & bc (top two) and load current in a & b phases (bottom
two) for a balanced 10 kW resistive load. Current is scaled at 1 A = 1 V.
Another test carried out to observe the dynamic response of the controller under
balanced conditions involved abrupt load change. Figure 611 demonstrates the response
load was abruptly connected using an threepole singlethrow switch. As seen in Figure
611, load terminal voltages on the secondary side of 480VΔ : 208 (120VY)
transformer respond fairly well with minor glitches at the instant of noload to load
switching. If it is assumed that the load is switched ON at t = tON, as the current is zero at
t = tON– the voltage is also zero; and as the ac current begins to flow through the load at t
= tON+ the voltage at load terminals regains its sinusoidal form. This voltage recovery
Figure 611 Experimental waveforms illustrating DR operation during transition from noload to a 2 kW
balanced threephase resistive. Load terminal lineline voltage ac & bc (top two) and load current in a & b
phases (bottom two). Current is scaled at 1 A = 1 V.
noload is illustrated Figure 612. As seen in this figure, the DR responds satisfactorily
Figure 612 Experimental waveforms illustrating DR operation during transition from a 2 kW balanced
threephase resistive load to noload. Load terminal lineline voltage ac & bc (top two) and load current in
a & b phases (bottom two). Current is scaled at 1 A = 1 V.
proved to satisfactorily regulate the terminal voltage under balanced load conditions. The
Inorder to test the performance of the controllers under unbalanced conditions, the
voltage waveforms at the load terminals across phases ac and bc as well as line currents
through phases a and b. Table 65 provides a comparison of the regulated load terminal
voltage with that of unregulated case. As seen in Table 65, the lineline voltages are
regulated fairly close to 208 V in all three phases. Small inconsistencies between
sensing circuits as well as in estimation of the leakage inductance of the second reactor
and transformer.
Figure 613 Experimental steadystate waveforms illustrating the load terminal voltage and load current
when load on phase a is abruptly disconnected. Current is scaled at 1 A = 1 V.
consisting of sequence filters PSF and NSF has a major impact on the effectiveness of
regulation of load terminal voltage. The functioning of PSF is demonstrated by the DSP
internal variables displayed in Figure 614 that are plotted from PC Master™ of
Figure 614 illustrates the filter input that is the transformer primary current, PSF output
in p.u.
in p.u.
in p.u.
Time (ms)
Figure 614 Transformer primary current, PSF output and NSF output waveforms in the PC Master™ for
an unbalanced load
As seen in Figure 614, the transformer primary current also contains harmonics
due to its magnetizing current. The sequence filters, PSF and NSF are employed to
extract the positive and negative sequence components of this distorted current, The PSF
and NSF have been designed to pass one sequence and reject the other at 60 Hz in an
efficient manner, but they do pass with some attenuation the higherorder harmonic
components. Nevertheless, Figure 614 demonstrates that the PSF and NSF satisfactorily
The next subsection deals with tests on external generation controllers in the
Chapter 4. To demonstrate the operation of the generation controls, tests on microgrid are
carried out in three modes, viz., (i) single DR in standalone mode, (ii) single DR in grid
interfaced mode and (iii) two DR interconnected mode. Selected results are presented
PL and PLref are plotted in Figure 615 and Figure 616, respectively. These waveforms
CodeWarrior™. The initial value of PLref is set at zero and therefore at a load of 0.25 p.u.
the frequency is 59.9375 Hz. As seen in Figure 615, a step change in PL from 0.25 p.u.
controller gain b) from the initial frequency of 59.9375 Hz and a steadystate deviation in
Time (ms)
Figure 615 Experimental waveforms showing response to a change in load real power demand PL when
the DR is operated as a standalone unit
180
Figure 616 illustrates the response for a change in PLref from 0 p.u. to 0.5 p.u.
that reflects the current load condition. As against the response to change in PL shown in
Figure 615, the step change in PLref gives a firstorder lag response with a timeconstant
Time (ms)
Figure 616 Experimental waveforms showing response to a change in load real power ref. setpoint PLref
to match the load demand when the DR is operated as a standalone unit
The grid frequency as observed from the point of common coupling (PCC) in the
laboratory experiment test bed was 59.96 Hz. However, the nominal frequency setpoint
of the power controllers programmed in the DSP is maintained at 60 Hz. The results
illustrating the response for a change in PLref from 0 p.u. to 0.5 p.u. are plotted in Figure
617. The load under this condition was 0.25 p.u.. Ideally, if the grid frequency were at
60 Hz, the generated power would have followed the PLref setpoint. However, as the
frequency is 59.96 Hz, the difference between PLref and generated power is determined
181
from the steadystate droop curves as (6059.96)/[b(1Kbβ)], which gives a value of 0.16
p.u. for PLref = 0 p.u. and to 0.66 p.u. for PLref = 0.5 p.u.. The DR frequency increases for
a short duration to enable higher generation but eventually returns to the grid frequency
of 59.96 Hz.
in p.u.
in Hz
in V
Time (ms)
Figure 617 Experimental waveforms showing response to a change in load power setpoint PLref from 0
p.u. to 0.5 p.u. when the DR is connected to grid interface PCC of frequency 59.96 Hz
Figure 618 illustrates the experimental waveforms when the load PL is changed
from 0.25 p.u.to 0.5 p.u.. As seen in this figure, the DR supplies only the transient change
in load but in steady state its generation returns to a value determined by PLref. Figure
619 illustrates the response for a change in PLref from 0.5 p.u. to 0 p.u.. As seen in this
figure, the DR generated power returns close to its earlier values when PLref is changed
back to zero.
182
Figure 618 Experimental waveforms illustrating the effect of load change at the DR terminals from 0.25
p.u. to 0.5 p.u. while it is operated in gridconnected mode. Load terminal voltage is the top waveform and
in the bottom are the load current (solid), DR current (dashed) and tieline current (dashdot). Voltage is at
100 V/div. and current is at 20 A/div.
in p.u..
in Hz
in V
Time (ms)
Figure 619 Experimental waveforms showing response to a change in load power setpoint PLref from 0.5
p.u. to 0 p.u. when the DR is connected to the grid interface PCC of frequency 59.96 Hz.
183
Figure 620 illustrates the results of DR1 response to a change in PL1 at its
terminals from 0.2 p.u. to 0.4 p.u.. The corresponding waveforms for DR2 that has a local
load of 0.4 p.u. are displayed in Figure 621. The load power setpoints in both units are
the same equal to 0.4 p.u.. Hence, as seen in these two figures the overall load is shared
equally by the two DRs. The final steadystate frequency is 60 Hz as the load burden of
Time (ms)
Figure 620 Experimental waveforms showing response of DR1 to a change in load PL1 from 0.2 p.u. to
0.4 p.u. when the two DRs are interconnected.
in p.u..
in Hz
in V
Time (ms)
Figure 621 Experimental waveforms showing response of DR2 to a change in load PL1 from 0.2 p.u. to
0.4 p.u. when the two DRs are interconnected.
184
Figure 622 illustrates the DR1 load terminal voltage, DR1 current and the tieline
current between the two DRs. As seen in the figure, since the two interconnected DRs are
supplying their local loads, the current through the intertie contains chiefly the
Figure 622 Experimental waveforms illustrating the two interconnected DRs supplying their local loads
with zero power flowing along the tieline. Load terminal voltage is the top waveform and in the bottom
are the DR current (solid) and tieline current (dashdot). Voltage is at 100 V/div. and current is at 20 A/div.
185
This thesis has presented control strategies for VSIbased distributed resources in
The concepts of feedback controller design for dc quantities using Bode plots of
loop gain have been extended for application towards threephase ac spacevector
quantities. These vector regulators have a PI controller cascaded with an oscillator. They
give a zero steadystate error when the oscillator frequency is equal to the excitation
frequency. The time response characteristics of peak overshoot and settling time for
vector regulated systems are evaluated against the frequencydomain parameters like the
phase margin and gain cross over frequency. When transformed to synchronous rotating
frames, these vector regulators are equivalent to the high performance complex vector
the processing involved that the synchronous frame based implementation of the vector
was applied to inverter based distributed resources to regulate the load terminal voltage
with zero steadystate error under unbalanced conditions. Frequency domain analysis
186
techniques in the form of Bode plots for individual sequence components were utilized in
the design procedure. Further, by making use of complex transfer functions, novel
negative and positive frequency/sequence selective filters that contain complex bandpass
as well as complex bandreject sections were proposed. The stability of complex filters
has been explained using Nyquist diagrams and Bode plots. The proposed
frequency/sequence selective complex filters give a high quality output without any
phase distortion. Simulation results were presented from digital simulation in Matlab®
reactive powervoltage controllers for the inverter based DRs was presented. These are
communication between the DRs. Smallsignal models were developed for microgrids
eigenvalue analysis and sufficient conditions were developed to ensure their smallsignal
stability. These eigenvalue investigations considered both the real and reactive power
flows in the microgrid. It was proved by means of mathematical propositions that the n
DR chain microgrid is stable for all positive values of generation controller gains with
the maximum droop limited to be less than the nominal value. Guidelines were provided
for design of the active powerfrequency and reactive powervoltage controllers. Besides
the results of dynamic behavior analysis, the IEEE P1547 performance specifications
187
[18] were also utilized in the design process. Simulation results are presented from digital
excursions of same quantity at the DRs that are connected to it. As a consequence, the
these excursions at the PCC. Preliminary results for stability of the parallel microgrid
were presented in the form of sufficient conditions. The preliminary results recommend
locating the smaller droop DR of higher generation capacity electrically closer to the
PCC making it the dominant DR in the parallel microgrid. Further, the frequency
restoration time constant (i.e. 1/ωG) of this dominant DR has to be large enough for easier
tracking of its frequency by the (smaller rated) DRs farther from the PCC. While the
violation of these sufficient conditions in a parallel microgrid has been shown to not
endanger a 5 DR microgrid, it remains to be proven when n DRs are present. The mesh
topology of microgrids was analyzed for the simplest case of a 3DR mesh network based
on its smallsignal analogy to the 3DR parallel network. As a nonzero R/X ratio is
typical of wires in medium voltage distribution systems, effects of R/X ratio were
7.1 Contributions
selected frequencies.
Regulator
for the filter inductor current and an outer loop for the filter capacitor
voltage is proposed.
189
Topologies
DRs connected in chain and parallel topologies. Considered both the real
of the excursions of the same quantities at the DRs that are connected to it.
• Analyzed the mesh topology of microgrids for the simplest case of a 3DR
network.
The research directions developed here may be extended further along the
individually.
gains. In this regard, the work earlier carried out on power systems
3. Field test the overall control scheme under realistic industrial grid
conditions including
Motor starting
192
Capacitor switching
Loss of load
Loss of line
7.3 Summary
the past several decades. However, the major focus was on motor drives and
uninterruptible power supplies (UPS). While these improvements have been substantial
in advancing the state of the art, they prove to be cumbersome for employing in power
system applications. The first part of this thesis addresses this issue by taking a systems
approach in dealing with power system issues such as imbalances and building vector
In the past, several schemes for active and reactive power controls for distributed
generation in a microgrid have been developed using a heuristic approach and were
models of an n DR microgrid. It is hoped that this work will provide motivation for
Bibliography
The Journal for Onsite Power Solutions, MarchApril 2004, pp. 89.
Semantic Hype or the Dawn of a New Era?,” IEEE Power and Energy Magazine, Vol. 1,
Generation,” IEEE Industry Applications Magazine, Vol. 7, No. 2, MarchApril 2001, pp.
8088.
England Electric,” IEEE Trans. on Energy Conversion, Vol. 10, No. 1, March 1995, pp.
194
169174.
[8] R. Lasseter, “MicroGrids,” Proc. of IEEE PES Winter Meeting 2002, Vol.
2002.
Proc. of IEEE PES Winter Meeting 2002, Vol. 1, pp. 315322, Jan. 2002.
Source Distributed Generation Systems,” Proc. of IASTED Power and Energy Systems
195
Networking Assets,” IEEE Spectrum Magazine, Vol. 38, No. 1, Jan 2001, pp. 8486, 88.
http://www.eere.energy.gov/distributedpower/research/ul_1741.html
Systems to Mitigate Load and Line Imbalances,” Proc. of IEEE PESC 2002, Vol. 4, pp.
Regulators Using Complex Vectors,” IEEE Trans. on Industry Applications, Vol. 36, No.
Connected ACDCAC Power Converter for Variable Speed Wind Energy Conversion
System,” IEEE APEC Conf. Record 2003, Vol. 1, pp. 154158, 2003.
196
for Fast FourQuadrant Drive Performance,” IEEE Trans. on Industry Applications, Vol.
Using the Complex Number Representation,” IEEE IAS Annual Meeting 1989 Record,
Controller with Capacitor Current Feedback and “BackEMF” Decoupling”, IEEE Power
for SinglePhase UPS Inverters,” IEEE Trans. on Industry Applications, Vol. 33, No. 2,
and Robust Control of UPS,” IEEE Trans. on Industrial Electronics, Vol. 48, No. 2,
PWM Inverters with Zero Steady State Error,” IEEE Trans. on Power Electronics, Vol.
Analysis of MultiLoop Voltage Regulation Strategies for Single and ThreePhase UPS
Systems,” IEEE Trans. on Power Electronics, Vol. 18, No. 5, Sept. 2003, pp. 11761185.
IEEE Trans. On Power Delivery, Vol. 16, No. 4, Oct. 2001, pp. 782790.
Distribution Systems,” IEEE Trans. on Industry Applications, Vol. 32, No. 6, Nov./Dec.
Generalized Integrators for Current Control of Active Power Filters With Zero Steady
State Error for Current Harmonics of Concern Under Unbalanced and Distorted
Active Filters,” IEEE Trans. on Power Electronics, Vol. 19, No. 2, March 2004, pp. 508
514.
Phasor Analysis in the Balanced Large Electric Power System,” IEEE Trans. on
Automatic Control, Vol. 40, No. 12, Nov. 1995, pp. 19751982.
Miranda, “Design of Fast and Robust Current Regulators for High Power Drives Based
on Complex State Variables,” IEEE IAS Annual Meeting 2003 Record, Salt Lake City,
Amsterdam, 1984.
Control for Power Quality Improvement of Distributed Generation,” IEEE 33rd Annual
Power Electronics Specialists Conference 2002, Vol. 4, pp. 1803 – 1808, June 2002.
[47] J. Liang, T.C. Green, G. Weiss, Q.C. Zhong, “Repetitive Control of Power
Conversion System from a Distributed Generator to the Utility Grid,” Proc. of the 2002
199
International Conference on Control Applications, 2002, Vol. 1, pp. 13 – 18, Sept. 2002.
[48] J. Liang, T.C. Green, G. Weiss, Q.C. Zhong, “Hybrid Control of Multiple
[49] G. Weiss, Q.C. Zhong, T.C. Green, J. Liang, “H∞ Repetitive Control of DC
AC Converters in Microgrids,” IEEE Trans. on Power Electronics, Vol. 19, No. 1, Jan.
Harmonic Isolator Using Active Series Filter," Proc. 4th Euro. Conf. on Power
Electronics and Appln. 1991, Florence, Italy, Vol. 3, pp. 030035, 1991.
Filter," IEEE Trans. on Power Electronics, Vol. 14, No. 2, March 1999, pp. 265272.
IEEE Trans. on Industry Applications, Vol. 39, No. 4, Jul.Aug. 2003, pp. 11091117.
200
[55] P. L. Jansen, The Integration of State Estimation, Control and Design for
Sharing Control,” IEEE Trans. on Industrial Electronics, Dec. 1990, pp. 506513.
Power Systems, John Wiley & Sons, Inc., New York, 1984.
Trans. on Industry Applications, Vol. 36, No. 1, Jan.Feb. 2000, pp. 131138.
2005.
Method of Analysis – Part 1," AIEE Transactions, July 1929, pp. 716730.
with Complex Coefficients,” IEEE Trans. on Audio and ElectroAcoustics, Vol. AU16,
Conf. on Circuit Theory and Design 1981, Hague, pp. 412419, 1981.
http://ewww.motorola.com/files/dsp/doc/user_guide/DSP56F805EVMUM.pdf
[76] Y. Suh, Analysis and Control of Three Phase AC/DC PWM Converter
Madison, 2004.
WisconsinMadison, 1997.
[78] H. Nyquist, “Regeneration Theory,” Bell System Tech. Journal, Vol. 11,
Systems Using Complex Transfer Functions,” IEEE PESC Conf. Record 1999, pp. 691
697, 1999.
[92] 1336 Plus V/Hz AC Drive User Manual, Rockwell Automation – Allen
systems with dynamic loads,” IEEE Trans. on Circuits and Systems I: Fundamental
Theory and Applications, Vol. 44, No. 9, Sept. 1997, pp. 796  812
power systems,” Proc. 1988 IEEE Int. Symp. Circuits Syst., pp. 905908, 1988.
[99] Draft Report for CERTS, Year 2000 Testing of Capstone and Honeywell
MTGs During Load Changes, Southern California Edison, Irvine, Feb 1, 2001.
[100] James Larmine, Andrew Dicks, Fuel Cell Systems Explained, John Wiley
Gates Energy Products, Inc., EDN Series for Design Engineers, ButterworthHeinemann,
MA, 1992.
[104] PNGV Battery Test Manual, U.S. Department of Energy, Idaho National
1999.
when operated in the gridinterfaced mode. This is made possible by locating them in
proximity with the critical/sensitive loads. During interruption in the grid power supply,
they need to seamlessly switch to the standalone mode and supply the total load demand.
contain large rotating masses in which energy is stored in the form of inertia. Under load
and supply transients, the inertia of such systems delivers or absorbs any power mismatch
between generation of prime mover and load demand. In contrast, inverter based DRs are
characterized by a low inertia. As a result, they do not have sufficient energy storage
capacity to meet the demands of loads in the event of transients. Thus, in order to provide
necessary to install auxiliary energy storage in them. If not, performance of the DRs may
Transient changes in load power demand may result from faults in transmission
line and/or load switching. For instance, a 75 kW Honeywell microturbine takes about
operation [99]. On the other hand, some fuel cells require about 10 s for a 15% change in
207
power output [100]. Moreover, a fuel cell also has a recovery period of a few minutes to
establish equilibrium before it can meet another step change in power output. The typical
response that can be expected of a microturbine for a step change in load demand is
illustrated in Figure A1. In the Figure, PL denotes the load power demand; PS is the
response of the microturbine and (PLPS) is the shortage in power that needs to be
supplied through some means. In gridconnected mode of operation, the grid supplies the
shortage in power until the microsource responds to a step change in power demand.
However, in standalone mode of operation this sudden demand can be met only if
Figure A1 Typical step response of a microturbine. PL  load power demand, PS  response of the micro
turbine
are achieved by the use of storage to provide a stable supply of power in the event of
208
possible when the DR with storage is controlled such that it supplies a particular
capacitors. Among these, the technology of batteries is the most developed and is well
established for a variety of applications. The other forms of energy storage are either still
in the prototype stage of development or are not suitable for mass production. Since a DR
link is the most appropriate location for installing storage (refer Figure 11). Thus,
battery storage in the dc link of the DR may be used to meet the power requirements
during abrupt changes in power demand from the load. The design issues in battery
The block schematic of a general battery based DR system is given in Figure A2,
in two possible configurations, (a) and (b). Figure A2(a) represents an unregulated dc
bus configuration and Figure A2(b) represents a regulated dc bus configuration. The
battery is represented in the figures as an ideal voltage source in series with an internal
resistance. Vbat is the terminal voltage of the battery, and PS, PL and Pbat are the micro
bus, and state of the battery during the charging and discharging cycles dictate the dc bus
voltage. Since the power flow in and out of the battery is for a very short duration, it
209
requires a battery with the ability to discharge to final voltage and recharge to 100% state
voltage is regulated and is independent of the battery charging and discharging cycles. A
dc/dc converter is used as the interface between the battery and the dc bus capacitor. The
dc bus voltage can be regulated by controlling the switching scheme of either the rectifier
The dc/dc converter in the regulated dc bus configuration may be used to step
down (buck) or stepup (boost) the voltage at the terminal of the battery. A battery
consists of several cells stacked in series, where a cell is its basic unit. It is generally
beneficial to have less current drawn from the battery, as this will maintain the
instantaneous power drawn from each cell to a minimum. So, it may be desirable to have
battery on the higher voltage side of the dc/dc converter. But the choice of too large a
battery voltage would require a large number of series connected cells resulting in charge
equalization problems.
rechargeable batteries include – leadacid, nickel cadmium (NiCd), nickel metal hydride
(NiMH) and lithium (Li)ion. NiCd and NiMH are not preferred in utility applications
since they have very high self discharge rate. Although, Liion has the highest energy
density, it is also the most expensive. The cost/kWh of a leadacid battery is the least
among these and it can meet the power density requirements of DG applications. So, a
leadacid battery is considered the most suitable for DG applications for further study in
this chapter.
cycle batteries are used in backup power applications where power is needed for longer
durations. These batteries are typically applied in electric vehicles and UPS systems.
Starting batteries are commonly used to start internal combustion engines. No power is
drawn from these batteries during the normal operation of the engines. They are capable
of providing large currents for a very short interval of time. Hence, starting batteries of
Battery sizing is dependent on the number of cells it contains, and the capacity of
each cell. The cell capacity is normally measured in amperehours denoted here as Ah. A
method of specifying charge and discharge rates of a cell is the C rate, which is defined
as the current flow rate that is numerically equal to the cell rated capacity [101]. For
example, C rate of a 1 Ah capacity cell is 1 ampere. Both charge and discharge rates are
normally represented as multiples of the C rate. A current that would nominally discharge
the battery in 1 hour is represented as 1C. A current that would nominally discharge the
discharge the battery in a short interval of 6 minutes would be represented as 10C. The
Leadacid batteries are usually rated at 8, 10 or 20 hour rates. For example, the
ABSOLYTE IIP series of valve regulated leadacid (VRLA) batteries are rated on a 8
hour rate, with capacities ranging from 105 to 4800 Ah [102], and NP series, rated on a
20 hour rate with capacities ranging from 1.2 to 65 Ah [103]. For a battery, the MSDG
rate of in these applications is about 36C. Battery parameters need to be derated at high
discharge rates, due to increased losses in it under these conditions. Table A1 gives the
nominal capacity of leadacid batteries at different rates of discharge. From the table, we
observe that a battery rated on a 10 hour rate has only 44% deliverable capacity if the
more common among battery manufacturers’ datasheets. Figure A3 shows the cell
provide tables containing dc amperes that the battery can source at different final volts
per cell and for different periods of time. These tables incorporate the derating necessary
for highrate discharge application for values at lower time intervals. It may be noted that
all battery ratings are for a nominal room temperature of 25oC or 77oF, and additional
Figure A3 Cell discharge voltage Vs. time for a leadacid battery [101]
In addition to loss of energy storage capacity, the terminal voltage of the battery
is also strongly affected by the discharge rate. Figure A4 illustrates the battery current
and the corresponding terminal voltage for two sizes of cell, ‘X’ and ‘D’, at two different
temperatures. In the figure, battery ‘X’ is rated at 3.2 Ah and ‘D’ at 1.8 Ah.
load can be drawn from the data represented in Figure A4. The negative slope of the
curves in Figure A4 gives the Thevenin resistance (Rth) of the battery. Using the
maximum power transfer theorem, we can deduce that, when the load equivalent
resistance equals Rth, maximum instantaneous power gets transmitted. Since Rth
213
corresponds to the battery’s internal resistance, we observe that at this operating point,
losses in the battery internal resistance are equal to the power transferred to load. Hence,
at the maximum power transfer operating point, the power dissipated in the battery would
be enormous. In order to avoid losses in the battery and subsequent deterioration in its
operation, it is necessary for us to select a battery whose internal resistance is very small
Figure A4 Peak current and voltage per cell of a leadacid battery for a highrate discharge [101]
drawn for ‘X’ and ‘D’ is shown in Figure A6. From the figure, it is clearly evident that
the battery has a limit on the maximum instantaneous power that it can support, which is
214
Figure A6 Instantaneous power curves of a leadacid battery for a highrate discharge [101]
when the load is suddenly removed and the microsource continues to supply power until
it is able to respond to the new power set point. The battery has to be capable of
accepting this difference in power for a short duration. Thus it must have a high charge
acceptance.
215
(HEVs). In this application, batteries are primarily used during starting, accelerating and
braking of the vehicle, and no net power is drawn from the battery in normal running
mode of the vehicle. Therefore, desirable attributes of highpower batteries for HEV
applications are highpeak power handling capability during starting and acceleration,
and high charge acceptance to maximize regenerative braking utilization. The basic
requirements for a power buffering battery in a HEV are defined by the Partnership for a
New Generation of Vehicles (PNGV) work group as shown in Table A2 [104].
requirements by optimal control of battery state using battery management systems [105].
In the near term, the general solution for most recently developed HEVs has been to
duration of the mismatch between the real power demand of the load and the power
delivered by the microsource. Since power drawn from the battery is for a short duration
capability are suitable. The main property of batteries for highdischarge applications is
their peak power capability, which in itself is a strong function of its capacity and internal
resistance. The construction of these batteries is optimized for peak power, i.e., high
capacity along with ultralow internal resistance. The batteries must also have a good
charge acceptance to accommodate the highrate charging upon removal of load. Most
state of the art batteries are hard pressed to meet these stringent requirements. Hence, the
design procedure is described in the form of a flowchart for sizing these batteries in
[10,15].
217
here as a “complex transfer function” and it is expressed in rectangular form with real
reference frame, the complex spacevector input quantity entering the filter is
quantities, the goal here is to verify that a complex transfer function may be used to
To verify the validity of this formulation, and to properly interpret the role of
function is expanded into real and imaginary parts. In particular, the inputoutput relation
complex transfer function and is illustrated in Figure B1. As seen in Figure B1, the
realization was proposed in [70] by Lang and Brackett. Thus, the twoinput twooutput
MIMO system of the form shown in Figure B1 is transformed to a singleinput single
N(s)
Gcpx(s) = D(s) (B8)
In the above equations, polynomials N*(s) and D*(s) denote the complex
xα(t) yα(t)
GRe(s)
GIm(s)
x(t) = xα(t) + jxβ(t) y(t) = yα(t) + jyβ(t)
GIm(s)
GRe(s)
xβ(t) yβ(t)
Figure B1 Realization of complex transfer function Gcpx(s) as a twoinput twooutput MIMO system
Substituting the above expressions for GRe(s) and GIm(s) in (B7), the transfer
function matrix of the MIMO system is determined in matrix fraction description (MFD)
form [106] as
⎡ GRe(s) –GIm(s) ⎤ 1
H(s) = ⎢ ⎥ = [Nr(s)]2x2 [Dr(s)]2x2 (B13)
⎣ GIm(s) GRe(s) ⎦
220
⎡ NRe(s) –NIm(s) ⎤
where [Nr(s)]2x2 = ⎢ ⎥ (B14)
⎣ NIm(s) NRe(s) ⎦
*
⎡ D(s) D (s) 0 ⎤
and [Dr(s)]2x2 = ⎢ * ⎥ (B15)
⎣ 0 D(s) D (s) ⎦
det([Dr(s)]2x2) = 0 (B16)
On the other hand, the poles of the complex transfer function Gcpx(s) describing
D(s) = 0 (B18)
system also include the conjugate pairs. Moreover, the MIMO system also displays
duplication of these complex conjugate pairs. Nevertheless, the two systems are observed
to have poles located in the same half plane of the complex splane. As a result, the
bounded input bounded output (BIBO) stability of MIMO system can be deduced from
the compact representation of the complex transfer function Gcpx(s) itself. The MIMO
system is BIBO stable if all the poles of Gcpx(s) lie in the left half of complex splane.
N(s) 1
Gcpx(s) = D(s) = (B19)
⎛  j ω60⎞
s
1+⎜ ⎟
⎝ ωo ⎠
N(s) = ωo (B20)
+ j ω60.
On the other hand, the transfer function matrix of the MIMO system is obtained
in MFD form as
1
H(s) = [Nr(s)]2x2 [Dr(s)]2x2 (B22)
2 2 1
⎡ ωo(s+ωo) ωoω60 ⎤ ⎡ (s+ωo) + ω60 0 ⎤
=⎢ ⎥⎢ ⎥ (B23)
⎣ ωoω60 ωo(s+ωo) ⎦ ⎣ 0 (s+ωo)2 + ω602 ⎦
⎡ A1 0 ⎤ ⎡ 0 1 ⎤
A=⎢ ⎥ where A1 = ⎢ ⎥ , (B24)
⎣ 0 A1 ⎦ ⎣ (ωo2 + ω602) 2ωo ⎦
0 0
⎡ω 0
⎤
B=⎢
0 ⎥
o
, (B25)
0
⎣0 ω ⎦
o
⎡ 0 1 ω60 0 ⎤
C=⎢ ⎥ and (B26)
⎣ ω60 0 0 1 ⎦
⎡0 0⎤
D=⎢ ⎥ (B27)
⎣0 0⎦
Matrix A contains two identical principal submatrices (labeled A1) that are
decoupled, and the eigenvalues of this submatrix A1 are determined to be ωo ± j ω60.
222
Therefore, eigenvalues of A are ωo ± j ω60 but repeated twice. The same is observed in
the case of poles of transfer function matrix H(s) that are the roots of
2 2
⎛ ⎡ (s+ωo) + ω60 0 ⎤⎞
det⎜ ⎢ 2 2
⎥⎟=0 (B28)
⎝⎣ 0 (s+ωo) + ω60 ⎦ ⎠
Thus, the poles of MIMO system include the conjugate pairs of the poles of
Gcpx(s), and these get duplicated too. As the two systems are observed to have poles
located in the same half plane of the complex splane, information on the BIBO stability
of MIMO system can be determined from that of the compact representation of complex
Lebih dari sekadar dokumen.
Temukan segala yang ditawarkan Scribd, termasuk buku dan buku audio dari penerbitpenerbit terkemuka.
Batalkan kapan saja.