Sunteți pe pagina 1din 9

948 IEEE Transactions on Power Systems, Vol. 10, No.

2, May 1995
CONCEPTS FOR. DESIGN OF FACTS CONTROLLERS
TO DAMP POWER SWINGS
Einar V. Larsen (F), Juan J. Sanchez-Gasca (M)
Power Systems Engineering Department
GE Industrial and Power Systems
Schenectady, NY 12345
Abstract - The design of controllers sited i n the transmission
network f or damping interarea power oscillations requires
several types of analytical tools and field verification methodr.
Probably the most important aspect of such control design is the
selection of proper feedback measurements from the network. This
paper describes concepts which provide design engineers with the
insight to control performance and the understanding needed to
ensure the secure operation of the bulk transmission system.
Specific attention is directed to procedures f or selecting feedback
signals.
Keywords - Power system stability, Flexible AC Transmission
Systems (FACTS), Control design, Modal decomposition
1. INTRODUCTION
A problem of interest in the power industry is the
mitigation of low frequency oscillations which often arise
between areas in a large interconnected power network
[l-31. These oscillations are due to the dynamics of
interarea power transfer and often exhibit poor damping
when the aggregate power transfer over a corridor is high
relative to the transmission strength. With utilities
increasing power exchanges over a fixed network, the use
of new equipment in the transmission systemto aid the
damping of these oscillations is being seriously considered.
A number of Static VAr Compensators (SVC) have already
been installed to aid power systemdynamics [4-61. The
Flexible AC Transmission Systems (FACTS) programof
EPRI is developing a number of new controllers for this
purpose, with Thyristor Controlled Series Compensation
(TCSC) now proven in the field and ready for
application [7-91.
Such controllers must operate satisfactorily in the
presence of many modes of power swings and over a wide
range of operation. Given the huge extent and variability of
the power system, the control design must be
I I
94 SM 532-2 PWRS
by the I EEE Power System mgi neeri ng Committee of the
IEEE Power Engineering Society for presentati on at
the IEEE/PES 1994 Summer Meeting, San Francisco, CA,
J ul y 24 - 28, 1994. Manuscript submitted December 30,
1993; made avai l abl e for pri nti ng J une 14, 1994.
A paper recommended and approved
I 1
Joe H. Chow (F)
Electrical, Computer & Systems Eng. Dept.
Rensselaer Polytechnic Institute
Troy, NY 12180-3590
understandable to engineers in many companies involved
witb the design, planning and operation of the power
system. Thus, insight to control performance is important to
ensure that the needed understanding is derived fromfield
measurements and fromsimulation results using a large-
scale simulation model.
Traditional approaches to aid the damping of power
swings include the use of Power SystemStabilizers (PSS)
to modulate the generator excitation control, for which
much experience and insight exist in the industry [10,11].
Unlike PSS control at a generator location, the speed
deviations of the machines of interest are not readily
available to a FACTS controller sited in the transmission
path. Further, since the intent is to damp complex swings
involving large numbers of generators, speed signals
themselves are not necessarily the best choice for an input
signal. For a FACTS controller, it is desired to extract an
input signal fromthe locally-measurable quantities at the
controller location. Finding an appropriate combination of
measurements is the most important aspect of control
design.
The main contribution of this paper is the presentation of
design concepts and a systematic approach for the
selection of input signals for FACTS damping controller
design. The approach is based on an approximate multi-
modal decomposition for systems with multiple swing
modes and a general input signal, essentially a
generalization of the well-known approach applied to a
single-machine model in the early days of PSS design
[10,11]. The approach takes advantage of a few
approximations to provide a computationally-efficient
means of displaying key information needed to understand
theimpact of input-signal selection and control design.
2. APPROXIMATE MULTI-MODAL
DECOMPOSITION
The approach described here makes a few
approximations to develop engineering insight in the
control design process. One key approximation is that the
modes of interest exhibit light damping. Another is that the
impact of any individual control on the frequency and mode
shape of the power swing is small. These assumptions
permit breaking the system apart based upon the
approximate mode shape information, and determining the
incremental effect of controllers on each mode separately.
0885-8950/95/$04.00 0 1994 IEEE
949
ith Modal System
- -
w
I . I
I'
Swing
Interest
. Mode of
Power
, System
(Except
ith Mode)
i
Effective Control Action
Figure 2.1. Multi-Modal Decomposition Block Diagram.
Theeffect of the controllers upon themselves also becomes
apparent, and the design can proceed using this information
to assure minimum self-interaction effects on the final
response.
2.1 Formulation
In the single-machine model [lo], the mechanical swing
mode is represented in terms of a synchronizing and a
damping torque with control loops built around it. In the
multi-modal decomposition, we pursue the same
representation for each of the swing modes.
Consider the linearized model of a multi-machine power
systemin the state space form
(2.1)
X=Ax+Bu, y=Cx+Du
where x is thevector of state variables, and U and y are the
vectors of control and measurement variables, respectively.
Weassume that in (2.1), the states are arranged in x as
x=[AS1 AS2 ... AS,Ao,~Ao,2 ... Am,, zT]'
(2.2)
where the AS's and the Aog's represent the generator
angles and speeds, respectively, and z is the vector of all
theother state variables. Thus the systemmatrices A and
B are in the form
(2.3)
where I is the identity matrix, o b is the systemfrequency
in radsec, and A21. The matrix A21 relates the generator
angles to the derivative of the speeds (acceleration), and
represents the network synchronizing effect normalized
with respect to the machine inertias. The machine
interactions and swing modes are largely determined by
A21 and A22, representing the synchronizing and damping
effects independent of any other state variables.
The multi-modal decomposition is a transformation based
on themodal decomposition of A21 using the eigenvectors
of A21. Thus it involves a relatively smaller amount of
computations since A21 is n x n where n is the number of
generators in the system. The full systemmatrix A can be
morethan 10 times larger than A21. Let V be the matrix of
the right eigenvectors of A21 such that V-1 A21 V =A,
where A is a diagonal matrix whose non-zero entries are
the normalized modal synchronizing Coefficients. Then the
transformation
v o o
x,=u-lx, U= 0 v 0
I. 0 j
(2.4)
applied to thesystem(2.1) results in
(2.5)
in =U-' AU x, +U-' BU =A, X, +B,U
y=CUx, +Du=C, X, +Du
The angles and speeds in the new system(2.5) represent
modal variables and thesystem. matrix A, has the structure
Theconcepts developed in this paper can be applied to
any subset of theswing-mode set. To illustrate the points,
subsequent discussion will focus on a single mode
evaluation. This approach can be used to evaluate one
modeat a time in any system, and may be adequate when
only one swing mode is dominant across a transmission
path.
For each swing mode hi with modal frequency Wi, the
state variables can be rearranged such that themodal angle
A6mi and speed AOmi corresponding to hi become thefnst
and second state variables, resulting in the system
representation
Fj] =[ !lcmi -d,,,i Ob - A d23] E j ] +[ ; ; ] U
Ad31 Ad32 Ad33
Y Cd2 Cd3] +Du (2.7)
[ : : I
where kmi and dmi are theapproximate modal synchronizing
and damping coefficients, respectively, and zmi consists of
all the other state variables. The modal frequency is
approximately given by ai =,/= o k . fad/SeC.
950
System (2.7) now resembles a single-mode system
considered in [lo]. As a result, we can construct a block
diagram to represent (2.7) using its transfer functions.
Using straightforward matrix manipulations, (2.7) is
expressed in the frequency domain
S AO~ ~ ( S ) =-(Ob/S)Kmi(S)A@mi(S) - Kd(S)U(s)
Y(s) = Koi (sh~mi (s) +KI L~ (SI U(S)
(2.8)
The forms of the various transfer functions are given in
Appendix A. A block diagramof (2.8) is shown in Figure
2.1. We denote Kmi(s). Kci(s), Koi(s) and KIL~(s) as the
modal, controllability, observability and inner loop transfer
functions, respectively. These transfer functions when
evaluated at s =j o, are complex, providing both gain and
phase information. This information can be used to select
the most effective transfer functions for feedback control.
Interpretations of Kmi(s). Kci(s), Koi(s) and KILi(S) are
given in subsequent sections.
The systemrepresentation (2.8) can also be obtained
from the complete eigenspace of A, rather than the
transformation (2.4). However, this approach is limited to
relatively small size systems [12] and one of the key
objectives here is to provide engineering insight in an
expeditious manner rather than exact results. At the
opposite extreme, only one mode shape need be known in
an approximate manner to usetheconcepts developed here.
2.2 Modal Characteristics and Control Influence
InFigure 2.1, themodal system involving thestates Asmi
and Aomi represents the lightly-damped swing mode. The
modal transfer function Kmi(s) includes the effect of the
electrical network, the generator fluxes, the controllers
such as the excitation systems and governors, and to a
much smaller extent, the other swing modes. This modal
transfer function will be predominantly real in the
frequency range of interest when analyzing the dominant
swing mode, and for s =joi, obKmi(joi) =mi2.
The efective control action (seeFigure 2.1)
(2.9)
describes the impact of a given damping controller;
KPSDC(S), on the modal system, and is useful in estimating
theeigenvalue sensitivity of the ith mode. FromAppendix
B, the perturbation Ah; =-AOi +jAwi of hi for a controller
KPSDC(S) with small gain is
(2.10)
AXi =- Kei ( joi)/2
that is,
Aoi =Real(Kei(joi)/2), Aoi =-Imag(Kei(joi)/2) (2.11)
This direct relationship of modal sensitivity to the
controller and power systemcharacteristics provides the
key to achieving the desired insight into control design.
Subsequent discussion will focus on each of the terms in
(2.9), to show how they relate to control performance and
how they can be used to select effective measurements.
Future analytical tool developments are expected to focus
onnumerically-efficient procedures to quantify these terms.
2.3 Interpretations
Controllability
The effect of a FACTS controller on a given swing mode
is defined as the controllability function Kci(S), and Kei(s)
is directly proportional to K~i(s). In PSS design, Kci(S)
depends mostly on the excitation system, generator flux
dynamics and network impedances. For network control
devices such as TCSC, Kci(s) depends mostly on the
network structure and loads. When evaluated at s =joi,
Kci(joi) provides a measure of how controllable the ith
mode is by the control signal U. If &i(jOi) is zero, then
modei is not affected by U. When morethan one damping
controller is used, Kci(s) is a vector transfer function.
In general, IGi(jOi) is different for different swing modes.
For example, a TCSC sited on a tie line would have
significant controllability on the associated inter-area
mode, but much smaller controllability over local modes.
In cases with multiple interarea modes, the controllability
may be nearly 180" out of phase fromone modeto the next.
Such a situation would mean that if the machine speeds
were averaged and transmitted to the controller, then the
action of improving the damping on one mode would
simultaneously decrease the damping of another mode.
This situation must be compensated for by using an
appropriate set of measurements so that this inherent
adverse impact will be minimized or eliminated.
The quantity Kci(jOi) is a good indicator for evaluating
effective locations to apply damping controllers, since the
larger this is the greater the leverage on the swing mode.
For example, an SVC located on a bus needing voltage
support will be more effective for damping control than one
close to a generator terminal bus.
Observability
The effective control action Kei(s) is also directly
proportional to the observability function Koi(s). The
function Koi(s) relates the measured signal y to the i t h
modal speed Aomi. For a PSS using the machine speed as
the input signal, Koi(s) =1 for the local mode of that
machine. When evaluated at s=j oi , Koi(joi) gives an
indication of the modal content of the ith swing modein the
measured signal y. Its magnitude can be used to assess the
effectiveness of measurements y for damping control
applications. In a multi-modal system, since Koi(S) is
defined with respect to the modal speed, measurements y
directly related to machine speeds will have observability
gains that are predominantly real. Signals more closely
related to angular separation, such as power flow, will have
an integral characteristic, i.e., nearly 90" of lag with
respect to the speed. If Koi(joi) is small, then the ith mode
is weakly observable fromthe measurement y. Thus having
large Koi(joi) for the dominant modes of interest is one of
the criteria in selecting an input signal for a damping
controller.
Inner Loop
The control design must also consider the effect of the
controller output on its input (i.e., the component of the
measured signal y that is due to the control U), other than
95 1
via the swing mode of interest. This effect may be
considered a feed-forward term, but here wehave called
it the inner-loop effect, symbolized by the transfer
function KILi(S). This choice is due to its impact on the
effective control action (2.9), as illustrated in Figure 2.1.
The inner loop transfer function KILi(s) is extremely
important in damping controller design using input signals
other than generator speeds. In Section 2.4, we develop
simple indices based on the constraint imposed by KIL~(s)
to aid the selection of an appropriate measurement signal.
2.4 Damping Controller Design Indices
In this section we define three indices which provide
insight to the performance of a damping controller with
given measurements. These indices are amendable to
efficient computation in a large system.
Controller Phase Index
When the controller gain is sufficiently small with
respect to the inner loop (i.e., IKILi(jOi) KPSDCuWi)I<<I),
(2.9) reduces to
a i -Kci(joi)KpsDC(joi)Koi(joi )/2 (2.12)
Wedefine the Controller Phase Index (CPI) for theith
CPI(i)=-(LK, (jo;) +LK,; (joi)) (2.13)
Then the phase of KPSDC(S) at s =joi should be set equal
to CPI(i) to achieve a pure damping influence.
The change in the swing mode damping is
approximately
modeas
where A$ =LKPSDC(jWi) - CPI(i) is the difference
between the actual phase and the pure damping phase for
thedamping controller.
Similarly, the change in the frequency of the swing mode
is
Ami =I Kps x (jwi]lKci (jwi]IKd(jwi)( sin(-A$)/2 (2.15)
For systems with several dominant swing modes, (2.14)
can be used to select control channels that are robust with
respect to the damping of these swing modes. Since power
system swing modes are usually in a small range of
frequencies, from0. 3 to 2 Hz, an ideal measurement signal
would have the property that the CPI remains roughly
constant and in the samequadrant over the frequency range
for all of the dominant swing modes.
Once a measurement signal is selected to satisfy this
property, selecting the time constants of lead-lag transfer
functions for KPS DC(S) is a relatively straightforward
exercise. From experience [l 11, matching the phase
perfectly is not as crucial as ensuring a good bandwidth, so
that higher-frequency interactions will not excessively limit
the gain. Some phase lag at the swing mode frequency is
acceptable, since it tends to increase the synchronizing
torques between the machines, consequently raising the
frequency.
Maximum Damping Influence Index
Referring to Figure 2.1, the controller KPSDC(S) forms a
feedback loop with KILi(S). For a robust design, we
constrain the gain on KPSDC(S) so as to maintain a gain
margin in the feedback system, beyond which the inner
loop will tend to dominate over the damping control. Using
a gain margin of 10 dB as a guide, the quality of a
measured signal can be quantified by a Maximum Dumping
Influence (MDI) index for each modehi
Tbe MDI index is a measure of the maximumeigenvalue
shift achievable assuming that the largest magnitude of the
damping controller is l/&lK,,,(jm,], where the & factor
guarantees a gain margin of 10 dB. Thus, given two
candidate measured signals yl and y2 with similar IKoi(jOi)l
for the dominant ith mode, y1 would be preferable if its
inner loop gain IKILi(jO)l is less than that of y2, because a
higher gain KPSDC(S) can be applied to provide more
damping enhancement to the ith mode.
The value of the MDI index is that it indicates the
effectiveness of measurements having high observability
gain and low inner loop gain. In Section 4, the MDI index
is used in a systematic way of scanning impedances for
synthesizing the bus voltage angular differences that are
effective for damping control design.
For systems with multiple controllers and measured
signals, KILi(S) is a matrix. In such cases, MDI may be
extended to provide information on potential adverse
controller interactions. This is a subject of further research.
Natural Phase Influence Index
In a practical design involving several operating
conditions, it is desirable to select a signal whose
observability will compensate for the change of the
controllability as systemcondition changes, such that a
fixed controller can be used to improve the damping of the
dominant swing modes. Such a property requires that the
CPI of the dominant modes for the different operating
conditions remain approximately the same. Thus, a robust
signal should have large MDI, and result in approximately
thesamephase requirement in compensator design over the
operating conditions. However, it may not always be
possible to find signals that meet these requirements in a
given system, especially when there are several dominant
modes. To address this situation, we develop an index
utilizing the impact of KILi(s) onthe systemzeros.
Much discussion exists in recent literature on the zeros
of thetotal closed-loop transfer function, fromthe control U
to the measurement y [2,13]. Frequency response
measurement of this transfer function in the field will show
both peaks and valleys, with the peaks corresponding to the
lightly damped poles, which are the swing modes hi, and
the valleys corresponding to the complex zeros zi of this
particular transfer function.
The complex zeros zi are the locations to which the
swing modes hi in the neighborhood of Zi will migrate in
the s-plane as the controller gain is increased to infinity,
which wecall the natural behavior of closing a loop with
952
the selected measurement signal. The initial trajectory of
this migration is influenced by the phase compensation
(2.12) of the controller. However, when the inner-loop
impact becomes too large the eigenvalues of the system
will tend toward these zeros. If Zi has less damping than
the corresponding hi, then the natural behavior of the
controller is to decrease the stability of hi - clearly an
undesirable situation. However, if Zi has more damping
than the corresponding hi, the natural behavior of the
controller will be to improve stability - clearly a good
situation.
To appreciate the relationship between KILi(s) and
systemzeros, welet IKpsx(s)I+m in (2.9), resulting in
Kei =-Kci Koi (~)/KI L~ (2.17)
Equation (2.17), which is inversely proportional to
KILi(s), represents the effective control action (2.9) with
infinite control gain. As IKPSDC(S)I + m, hi + Zi. Using
the eigenvalue sensitivity (2.12), Ahni , the "natural"
change in the ith swing mode hi due to the controller, is
approximately
Ahni =zi -hi =-Kci(jwi)Koi(joi)/(2K1~(jcoi)) (2.18)
Note that this "natural" change will have magnitude and
phase characteristics. The magnitude aspect is already
captured by the MDI index of (2.16). The phase-related
index is quantified by a Natural Phase Influence (NPI) index
for each modehi
WI (i)= -(CPI(i)+LKIL (jcoi)) (2.19)
A value of NPI(i) =-90" means a pure positive
synchronizing influence, while NPI(i) =0" implies a pure
positive damping influence. Thus, the selection of the
measured signals should favor those which have NPI(i)
near 0". Webelieve the best region is for NPI(i) to lie in
the quadrant O">NPI(i)>-90", for which the natural
tendency is toward improving both damping and
synchronizing torques.
3. TESTSYSTEM
We will illustrate the design concepts and the signal
selection process using the 2-area, 4-machine system
proposed in [141 (Figure 3.1). Each area consists of two
machines, one small and one large to represent a grouping.
W - E Power Transfer
Local Local
____)
TCSC Mode2
-i XI,
900 MVA
900 MVA
p.165
p.165
1550 MVA 1350 MVA
Figure 3. I . Two-Area Four-Muchine System
The system data are the same as those given in [14],
except that the machine MVA ratings and the impedances
of the tie lines between the coherent areas have been
changed (as listed in Figure 3.1) and a TCSC is included in
one of the tie lines. The systemexhibits an interarea mode
near 0.6 Hz and two local modes between the machines
within in each area, with damping as noted in Table 3.1.
Thebase systemhas a total of 600 MW power transfer on
three tie lines, while the contingency systemhas the same
power transfer on two tie lines, with tie line 3 not
in-service.
TABLE 3.1
SWING MODE FREQUENCY AND DAMPING
Interuea Mode Local Mode 1 Local Mode 2
Base 0.62 Hz, 0.17% 1.15 Hz, 11% 1.02 Hz, 18%
Contingency 0.52 Hz, -0.13% 1.13 Hz, loa0 1.00 Hz, 18%
4. MEASUREMENT SELECTION
4.1 Measurement Synthesis
The TCSC control for damping of the interarea mode
uses locally-available measurements to synthesize the
controller input signals. The synthesized signal should be
sensitive to the interarea mode of oscillation, while
insensitive to the local modes. In addition, the useof the
signals should not cause any adverse interactions between
the controllers.
In previous TCSC and TCPR (thyristor controlled phase
regulator) controller desigd investigations [7,8,9,15], the
angular difference of synthesized remote voltages on each
side of the control has shown promise as a good
measurement for damping control. This concept is
illustrated in Figure 4.1. By measuring the through current
IT and thevoltages Vmeasi and Vmeas2 at the terminals of a
TCSC, the remote complex voltages Vs ynl and Vsyn2 can
be synthesized as a function of the synthesizing
impedances Zsynl and Z3yn2:
Thesynthesized angular difference is
012 =4 y n l - 4 y n 2 (4.2)
For a relatively wide range of Zs ynt and Z s y n ~ , the
synthesized 012 will consist of mostly the interarea mode
with an insignificant amount of the local modes, that is, 012
will mimic the angular difference between the center of
machines of the two areas. However, the impact of the
TCSC action on this measurement (the inner-loop effect)
will bemore sensitive to the selection of Zs ynl and Z s y n ~ .
To a first approximation, setting these impedances close to
Figure 4. I . Synthesized Voltages and Angular Difference.
953
the driving-point impedance seen in both directions is a
good starting point. In Section 4.2, we apply the multi-
modal decomposition for a systematic selection of Zs ynl
and Z s y n ~ , using the MDI and NPI indices described in
Section 2.4.
4.2 Synthesizing Impedance Selection
To search for the appropriate Zsynl and Zsyn2, wescan the
values of the MDI and NPI indices for a range of Zsynl and
Z s y d values. The simulation shown here uses
Zspi = +jXsyni, Zsyn2 =Rsyn2 +j X s y n ~ (4.3)
where the resistances R y n l and Rsyn2 are set at 10% of the
reactances Xspi and Xsyn2, respectively.
Figure 4.2 is a three-dimensional plot showing the MDI
indices in the '2'-dimension over a ( Xsynl , Xsyn2) grid. The
plot shows the desirable region where the MDI indices are
large, essentially indicating an inverse relationship of the
two synthesizing reactances is desired for best damping
performance (e.g., if one is large, the other should besmall,
and visa-versa).
Figure 4.3 is a three-dimensional plot showing the NPI
indices in the '2'-dimension over the ( Xsynl , Xsyn2) grid.
Figure 4.2. MDI Indices -vs- Synthesizing Reactances for
the Interarea Mode of the Base System
0 5
-.
0.5 0
A -'\
I
\ Contingency System
I
\r-l MD" l
L
n
0 1.0 0.2 0.3
Xsy"l* P"
0.4
Figure4.4. "Good" MDI Regions for Base and
Note that this has a ridge of similar shape as the MDI
index, with the phase being near 270" (=-90") on the side
of smaller synthesizing reactances and +90" for larger
reactances. Given our desire to keep closer to the former
angle, any deviation from the high-MDI synthesizing
reactances should be on the low side for as many cases as
possible.
Figure 4.4 illustrates the regions of MDb1 for the base
and contingency systems. Note that the region of MDbl
for the contingency case is wider and at higher synthesizing
reactances than for the.base case. The trend toward higher
reactances is due to the loss of a portion of the transmission
system, thereby increasing impedance as seen fromany
bus. The wider region of high MDI is due to the
contingency removing one of the parallel paths to the line
having the TCSC; thus, the TCSC has more leverage on
theintertie power swings.
Contingency Cases.
5. TCSC DAMPING CONTROL PERFORMANCE
To illustrate the previous points, a damping control will
be applied with three choices of synthesizing reactances.
These choices are labeled as Points A, B and C in Figure
4.4, selected to show optimum and adjacent points for the
base case. Point A has an MDI of 0.3, with an NPI on the
"good" side (e.g., -90" for improved synchronizing torque).
Point C also has an MDI of 0.3, but with an NPI on the
"bad" side (e.g., +90" which decreases synchronizing
torque). Point B is onthe peak of the MDI characteristic.
Based on the CPI =-90" for the interarea mode, the
damping controller
1 S S
KPSDC(S)=-K~---
1+0.04s l +s 1+0.04s
Figure 4.3. NPI Indices -vs- Synthesizing Reactances for
the Interarea Mode of the Base System.
954
5.5 t
1
1
5 t Zero (A)
I
x x
2.5 1
i
Zero (C)
2 1 I I I I
I I I
-1 -0.8 -.06 -.04 6.2 0 0.2 0.4
Real Axis
Figure 5.1. Root Loci of the Interarea Mode for Damping
Controllers Using Measurements A, S, or C.
is used, where [s/(l +O.O4s)][l/(l +0.04s)I is the filtered
derivative circuit to convert the angle measurement into a
speed measurement, and s/(l +s) is a washout circuit.
A
root locus analysis on& is shown in Figure 5.1, for each of
the three choices of synthesizing impedance.
For synthesis choice B, the controller provides
considerable damping with no curl in the locus evident in
this scale. For the other two choices, the root loci curve
and approach their respective zeros. As expected fromthe
NPI indices, choice A exhibits an increase in frequency,
whilechoice C exhibits a decrease in frequency. Also, the
improvement in damping of these non-optimumchoices is
limited to approximately 0.3 before the effect of the zero
becomes significant, as predicted by the MDI index.
6. CONCLUSIONS
The concepts described in this paper can be developed
into a set of analytical tools which will provide key insights
to aid the task of designing FACTS controllers to damp
interarea power oscillations. Several important aspects of
controller performance are explained, and quantifiable
indices are defined which will permit the evaluation of a
large number of operating conditions and contingencies in
anexpeditious manner. Such a tool is expected to augment
existing dynamic simulation methods, so that design
engineers can develop a good understanding of the results
seen in full-systemsimulations and can specify field tests
to verify the essential aspects of control performance.
7. ACKNOWLEDGMENTS
The support for this work has been provided by EPRI,
recently under RP3022-2 plus early development under
RP2707-1, and by NSF under grant ECS-9215076.
J.H.Chow has also been supported by an RPI sabbatical
leave at GE. The support and encouragement of Stig
Nilsson, Neal Balu, and Dominic Maratukulamat EPRI,
and John Hauer at BPA is greatly appreciated. John
Paserba and Ann Hill at GE and Glauco Taranto at RPI
havecontributed with analysis of example caqes.
8. REFERENCES
J.F. Hauer, Robust Damping Controls for Large
Power Systems, IEEE Controls Systems Mag., pp. 12-
.18. Jan. 1989.
J.F. Hauer, Reactive Power Control as a Means of
Enhanced Interarea Damping in the Western Power
System - A Frequency Domain Perspective
Considering Robustness Needs, in Applications of
Static Var Systems for System Dynamic Pevormance,
IEEE Publication 87TH01875-5-PWR, 1987,
J.H Chow, Ed., Time-Scale Modeling of Dynamic
Networks with Applications to Power System, Lecture
Notes in Control and Information Sciences, Vol. 46,
Springer-Verlag, 1982.
RJ . Piwko, Ed., Applications of Static Var Systems for
System Dynamic Performance, IEEE Publication
E.V. Larsen, N. Rostamkolai, D.A. Fisher and
A.E. Poitras, Design of a Supplementary Modulation
Control Function for the Chester SVC, Trans. Power
General Electric Co., Improved Static Var
Compensator Control, EPRI Report TR-100696, Final
Report for Project RP2707-1, June 1992.
General Electric Co., Flexible AC Transmission
Systems (FACTS): Scoping Study, Vol. 2, Part 1:
Analytical Studies, EPRI Report EL-6943, Final
Report for Project RP3022-2, September 1991.
W. A. Mittelstadt, B. Furumatsu, P. Ferron J . Paserba,
Planning and Testing for Thyristor Controlled Series
Capacitors, In Current Activity in Flexible AC
Transmission Systems, IEEE Publication 92 TH 0465-5
PWR. April1992.
S . Nyati, C.A. Wegner, R.W. Delmerico, R.J. Piwko,
D.H. Baker, A. Edris, Effectiveness of Thyrisfor
Controlled Series Capacitor in Enhancing Power
System Dynamics: An Analog Simulator Study, paper
93 SM 432-5 PWRD, IEEWPES Summer Meeting,
Vancouver, B.C., July 18-22, 1993.
87TH01875-5-PWR, 1987.
Deliv., Vol. 8, pp. 719-724, April 1993.
[lo] F.P. DeMello and C. Concordia, Concepts of
Synchronous Machine Stability as Affected by
Excitation Control, IEEE Transactions on Power
Apparatus and Systems, Vol. PAS-88, pp. 316-329,
1%9.
[ll] E.V. &en and D.A. Swann, Applying Power System
Stabilizers, Parts I - 111, IEEE Transactions on Power
Apparatus and Systems, Vol. PAS-100, pp. 3017-3046,
1981.
[12] P. Kundur, G.J. Rogers, D.Y. Wong, L. Wang and
M.G. Lauby, A Comprehensive Computer Program
package for Small Signal Stability Analysis of Power
Systems, IEEE Trans. on Power Systems, Vol. 5, pp.
[13] N. Martins and L.T. G. Lima, Eigenvalue and
Frequency Domain Analysis of Small-Signal
Electromechanical Stability Problems, in
Applications of Static Var Systems for System Dynamic
Performance, IEEE Publication 87TH01875-5-PWR,
1987.
1076-1083, 1990.
955
M. Klein, GJ. Rogers and P. Kundur, A Fundamental
Study of Inter-Area Oscillations in Power Systems,
IEEE Trans. on Power Systems, Vol. 6, pp. 914-920,
1991.
A.T. Hill, et al. Flexible AC Transmission Systems
(FACTS): System Studies - Thyristor Controller
Retrofit t o an Existing Western Area Power
Administration Phase Shifter, Final Report, EPRI
RP3022-02, October 1993.
J.H. Chow and J.J. Sanchez-Gasca, Pole-Placement
Design of Power SystemStabilizers. IEEE Trans. on
Power Systems, Vol. 4, pp. 271-277, 1989.
Appendix A
TRANSFER FUNCTIONS IN FIGURE 2.1
The modal controllability &i(S) is the transfer function
fromU to A h h and is given by
Kci(s)=Ad23(S1-Ad33)-1 Bd3 +Bd2 (A. 1 )
The modal observability of a measurement Koi(s) is the
transfer function fromthemodal speed to themeasurement
y, and is computed as
Theinner loop gain is given by the transfer function from
thecontrol U to the measurement y, with no change in Wmi
or &. It is computed as
(A.3)
There is also a transfer function fromthe modal angle to
themodal torque and it is given by
KI L (s)=C~~(SI - Ad33)-l Bd3 +D
[ Ad31 + S Ad32)
(A.4)
Appendix B
DERIVATION OF (2.10)
We use an analysis similar to that contained in [16] to
show (2.10). The characteristic polynomial of the closed-
loop systemwith thecontroller KPSDC(S) is
S2 +Kei (S) S +a b Kmi (S) =O
(B.1)
Wewill neglect the real part of hi such that hi =j o i and
assume obK mi (i oi ) J Oi 2. Let Kei QUi ) =Kre +j Ki m, and
theroots of (B.l) be s =- AOi +j ( 0i +AOi). Substituting
these quantities into (B.l), weobtain
Multiplying out the terms in (B.2) and neglecting the
higher order perturbation terms, (B.2) reduces to two
equations, one for the real part and the other for the
imaginary part
2A0i0i +Ki rnwi J 0, -2A0i0 +Krewi J 0 (B.3)
BIOGRAPHY
Ei nu V. Lalaea (F91) BSEE 73, Cal Poly, San Luis Obispo,
c.liforni., MSEPE 74 Rensselaer Polytechnic Institute, Troy, New York.
Mr. Lusen has been with GE since graduation in 1974. He has spent
mart of his career in GEs Power Systems Engineering Department in
Scbenectuly, New York, with afew yws in GEs HVDC a d SVC product
deputmnt in Philadelphia. He has been involved in many aspects of
controls for pows system devices, including generator exciters, HVDC
systems, SVC system, and power plant steamsupply. He developed many
new digital-computer tools for analyzing power system dynamics, and led
several field tests. He has also been involved with harmonics on power
systems, and filter design.
Mr. Lamen is active i n IEEE and CIGRE, currently chairing the IEEE
Working Group on FACTS.
Juan J. SanchaCasea received his PhD in Electrical Engineering
from theUniversity of Wisconsin-Madison in 1983. ?hat year he joined GE
where he is a Senior Application Engineer in the Power Systems Engineering
Department. He is interested in thedynamic simulation and control of power
systems.
Joc H. Chow (F92) received his Ph.D. degree in Electrical
Engineering fromthe University of Illinois in Urbana in 1977. From 1978 to
1987 and in 1993 (sabbatical leave) he worked at General Electric Company
in Schenectady. He joined Rensselaer Polytechnic Institute in 1987 and is
currently Professor of Electrical, Computer and Systems Engineering, and
Electric Power Engineering. His current research interests include power
systemdynamics and control, large scale systems and robust control.
(-Ao~ +j (q +Ami)) 2 +(Kre +jKim)(-Ao, +j (Oi +Ami))
956
DISCUSSION E.V. Lawn, JJ. Sanchez-Gasca, J.H. Chow: We wish to
D. J. TRUDNOWSKI (Pacific Northwest Laboratory, Richland, thank Dr. Trudnowski for an opportunity to clarify the
Washington): The authors are congratulated on providing a valuable
perspective on selecting signals for feedback damping control. The assumptions made in deriving (2.19). The
potential contributions go beyond power systemapplications to
include many large-scale dynamic systems. approximation (2.10) holds when &i(joi) is small compared
There is some confusion in the derivation of NPI index in equation
(2.19) which is derived from(2.18). Equation (2.18) is developed
by combining (2.17) and (2.101, but it seems that one cannot use
(2.17) and (2.10) together. This is becuase (2.10) is derived
assuming a small controller gain, and (2.17) is developed
assuming the controller gain goes to infinity. One way to aleviate
this problemis to remove the small gain restriction in deriving
(2.10). Could theauthors comment on this issue.
Manuscript received August 19, 1994.
to of [see equation @.I)]. ~f the controllef gain is small,
this assumption will be satisfied. The natural phase index
PI is useful when the mt-locus branch from the lightly
damped pole terminates on a near-by zero. In this case,
&i(jq) will be small, and equation (2.19) is valid.
Manuscript received October 26, 1994.

S-ar putea să vă placă și