See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/4126238

A new fuzzy logic power system stabilizer performances

Conference Paper · November 2004

DOI: 10.1109/PSCE.2004.1397498 · Source: IEEE Xplore

CITATIONS READS
19 3,560

2 authors:

Manuela Dobrescu Innocent Kamwa

Hydro-Québec Laval University
9 PUBLICATIONS   123 CITATIONS    410 PUBLICATIONS   12,062 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Hydro Quebec Strategic Power System Defense Program View project

Rapid Design Method for Generating Power System Stability Databases in SPS for Machine Learning View project

All content following this page was uploaded by Innocent Kamwa on 13 May 2014.

The user has requested enhancement of the downloaded file.

 1

A New Fuzzy Logic Power System Stabilizer

Performances
M. Dobrescu, Non Member, and I. Kamwa, Senior Member, IEEE

either due to switching actions in the short term or system

Abstract-- This paper describes the performances of a new enhancements in the long term.
power system stabilizer, the Fuzzy Logic PSS (FLPSS). It is
basically a PID (proportional-integral-derivative) type FLPSS With the conventional fixed-parameters stabilizers, the
with adjustable gains added outside in order to keep a simple gains and other parameters may not ideally suit the entire
structure. The FLPSS uses the generator speed deviation as spectrum of operation. The main objective of PSS is to insure
primary input from which the accelerating power is derived as a system stability and good performances for all operating
secondary input. In order to validate the FLPSS, it has been conditions and network configurations.
compared with two reference stabilizers, the IEEE PSS4B and
IEEE PSS2B from the IEEE Std 421.5. Conclusions are Nowadays, with the development in digital technology, it
supported by a range of small and large signal analyses, has become possible to develop and implement new
performed on a four machine two areas test system (with two controllers based on modern and more sophisticated synthesis
configurations). techniques. Indeed, controllers based on robust optimal
control, adaptive control, artificial intelligence are being
Index Terms-- Electromechanical modes, fuzzy logic, large developed. Among these methods, fuzzy logic is particularly
signal stability, power system stabilizer, small signal stability. attractive because it does not require a mathematical system
model to be controlled. It is, therefore, well-suited when the
I. INTRODUCTION system to be controlled is complex, nonlinear and difficult to

I N the context of modern interconnected electrical

networks, power systems stabilizers are considered an
essential control mean to improve stability and transmission
model. This powerful tool was used in different fields of
application including, recently, power systems dynamic
performance.
capacity. Extensive network interconnections tend to The Fuzzy Logic PSS concept was derived to provide an
introduce new electromechanical modes of oscillation original and efficient solution to this wide range modes
between electrically coherent power plants or areas. When damping problem. The core of the paper is divided in three
large electrical areas are involved, corresponding inter-area sections, one to describe the FLPSS itself, a second one to
modes may be as low as 0.2 Hz. However, inter-machines bring some results based on small signal stability simulations
oscillations inside a given power plant may reach frequencies and a third one to present some large signal stability results.
as high as 4.0 Hz when machine inertia are small and exciter In order to emphasize the FLPSS performances we will
gains are high. Power system stabilizers are facing such a compare this novel fuzzy logic PSS results with two reference
wide range of oscillating modes that they must ideally damp stabilizers results: the IEEE PSS2B [7] and IEEE PSS4B [1].
efficiently.
The IEEE PSS2B is an integral of generator accelerating
Most power system stabilizers in use in electric power power using a three-stage lead-lag transfer function while the
systems are derived from the classical linear control theory. IEEE PSS4B is built on a flexible multi-band transfer function
This theory is based on a linear model of a fixed power structure to provide more degrees-of-freedom for achieving a
system configuration. In other words, a fixed-parameters robust PSS tuning over a wide frequency range. Both of them
power system stabilizer, called a conventional PSS is are included in the new IEEE Std-421.5 [2]. These IEEE PSS
optimum for one set of operating conditions and may not be have the same external inputs (speed and electrical power
as effective for drastically different set of operating deviations). Those models and parameters are presented in the
conditions and/or network configurations. Appendix.
Indeed, power systems are nonlinear systems and their The reference system data and contingencies analysis, the
operation is basically of a stochastic nature. Therefore, system PSS structures and settings used in this paper were
configuration is dynamic with frequent topological changes implemented in Matlab/SimPowerSystems software [8].

II. DESCRIPTION OF THE FLPSS

M. Dobrescu and I. Kamwa are with Hydro-Québec, IREQ, 1800 Boul. The FLPSS model is shown in Fig.1. As for a conventional
Lionel-Boulet, Varennes, Quebec, Canada, J3X 1S1 (e-mail: PSS, the FLPSS comprises three main blocks, the filters
[email protected], [email protected])
 2

(WO), the fuzzy logic controller (FLC) and the limiter. One The output variable is classified by triangular fuzzy
adjustable limiter (±0.15) is provided for the FLPSS output. membership functions: BN (Big Negative) = [ -1.0 -0.51
The filters WO are two washout filters with time constants of -0.439], MN(Medium Negative)= [-0.515 -0.439 -0.302],
Tw=10 seconds. LN(Low Negative)= [-0.439 -0.302 0.0], BZ (Big Zero) =
[-0.302 0.0 0.302], LZ(Low Zero)= [-0.0015 0.0 0.0015],
It is, as we can see, a PID (proportional-integral-
LP(Low Positive)=[0.0 0.302 0.439], MP(Medium
derivative) type stabilizer [6] which uses the speed deviation
of the machine as primary input. The core of the FLPSS is a Positive)= [0.302 0.439 0.515], BP(Big Positive)= [0.439
fuzzy logic controller (FLC). Gains have been added outside 0.515 1.0].
the FLC, on the one hand, to normalize input variables of the The output signal was obtained using the following
fuzzy logic controller and, on the other hand, to achieve the principles:
PID function. Let's note that the addition of these gains - If the speed deviation is important, but tends to decrease,
outside fuzzy logic controller permits to keep a simple then the control must be moderated. In other words, when
structure. The FLC uses the speed and the acceleration power the machine decelerates, even though the speed is
deviations as inputs, the latter being derived from the speed. important, the system is capable, by itself, to return to
∆ω PD
steady state.
VS - If the speed deviation is weak, but tends to increase, then
WO Kw FLC Ks1 the control must be significant. In this case, it means that,
if the machine accelerates, the control must permit to
∆Pa Σ reverse the situation.

2Hs Kp The inference mechanism of the FLC is represented by a

7x7 decision table. The entire set of rules (49 if-then rules) is
s-1 Ks2 presented in Table I.
TABLE I.
PI FLPSS DECISION TABLE
Fig.1. PID type FLPSS model
∆w/
In what follows, we will describe how the FLPSS has been BP MP LP Z LN MN BN
∆Pa
synthesised. In a first place, we are going to show how the
fuzzy logic controller was realised from concepts of fuzzy BN BZ LN MN MN BN BN BN
logic [5]. Thereafter, we will present the tuning gains MN LP BZ LN MN MN BN BN
methodology. LN MP LP BZ LN LN MN BN
A. Fuzzy logic controller Z BP MP LP LZ LN MN BN
The design process of the fuzzy logic controller may be LP BP MP LP LP BZ LN MN
split into five steps: the selection of control variables, the MP BP BP MP MP LP BZ LN
membership function definition, the rule creation, the fuzzy
BP BP BP BP MP MP LP BZ
inference and the defuzzification strategy.
The fuzzy inference method is minimum-maximum type B. Tuning FLPSS gains strategy
(Mamdani). The defuzzification strategy used is the fuzzy The FLPSS was obtained by combining the fuzzy logic
centroid method. controller with proportional-derivative action (FL-PD) and the
The two input variables membership functions have been fuzzy logic controller with proportional-integral action (FL-
chosen identical because of the normalization achieved on the PI). The gains of the proportional, derivative and integral
physical variables. The normalization is important because it actions of the FLPSS are given by the following relations:
allows the controller to associate an equitable weight to each K PR = Ks 2 × F {Kp} + Ks1 × F {Kw}
of the rules and, therefore, to calculate correctly the stability
signal. K INT = Ks 2 × F {Kw}
(1)
K DER = Ks1 × F {Kp}
Each of the input variables is classified by seven
trapezoidal fuzzy membership functions. The following fuzzy Vs PID = K PR × ∆ω + K INT × ∫ ∆ωdt + K DER × Pa
sets were chosen: BN(Big Negative)= [-5.0 –1.0 –0.2 0.1],
MN(Medium Negative)= [-1.0 -0.2 -0.1 0.2], LN(Low In (1) we used the following notations:
Negative)= [-1.0 -0.1 -0.03 1.0], Z(Zero)= [-1.0 -0.03 0.03 - KPR is the proportional action gain.
1.0], LP(Low Positive)=[-1.0 -0.03 0.1 1.0], MP(Medium - KINT is the integral action gain.
Positive)= [-0.2 0.1 0.2 1.0] and BP(Big Positive)= [-0.1 0.2 - KDER is the derivative action gain.
1.0 5.0]. - The F{} term represents the fuzzy operation.
 3

For obtaining the gains Kw, Kp, Ks1 and Ks2 a three steps A. Small signal analysis
method has been used. This consists in adjusting only one
parameter at a time and is designed with the following steps: Small-signal analysis provides a mean to compare the
damping of the different system modes. The values were
1. Normalize input variables (the adjustment of Kw and Kp
obtained with no stabilizer at either site first and then with
gains).
FLPSS, IEEE PSS4B and IEEE PSS2B in closed-loop at the
2. Tune the FL-PD control first without using PI control. four generators in Fig.2. In order to identify the open-loop
3. Keep input gains Kw and Kp unchanged after adding FL- system, a finite impulse (1% from reference voltage Vref for
PI control. Adjust the output gains Ks1 and Ks2 in FL-PI 0.2 seconds) is injected in turn into each of the four machines
(proportional-integral) and FL-PD (proportional- while recording the output response signals. The linear MIMO
derivative) branches to obtain a good result. model is constructed from a modal analysis of these time
responses [9]. It allows the dominant modes to be identified.
At the end, the following values were obtained for the
FLPSS gains: Kp=0.55, Kw=50*Kp, Ks1=1.2 and Ks2=0.12. To better understand the results, we have completed the
small signal analysis by providing in Table II and Table III the
III. RESULTS BASED ON SIMULATIONS modal performance of the three PSS on a single tie-line
We will start comparing the three PSS using a four system (K1L) and on two tie-lines system (K2L). This four
machine two areas system called in this paper the Kundur test machines two areas system has a strong inter-area mode.
system [3-4]. In its basic symmetrical configuration, this TABLE II.
system is available in the Matlab/SimPowerSystems software PSS DAMPING ON A SINGLE TIE LINE SYSTEM
[8] as a demonstration. Despite its small size, it mimics very
Mode Open Loop FLPSS IEEE PSS4B IEEE PSS2B
closely the behavior of typical systems in actual operation. Hz ζ Hz ζ Hz ζ Hz ζ
We have implemented the stabilizers: FLPSS (described in Inter- 0.44 -0.015 0.37 0.26 0.33 0.37 0.46 0.19
area
previous section), IEEE PSS2B and IEEE PSS4B (described Local 1.10 0.12 0.83 0.59 1.87 0.5 1.04 0.33
in Appendix) on all generating units (G1, G2, G3 and G4). area 1
Local 1.15 0.095 0.86 0.61 1.99 0.45 1.07 0.32
The Kundur test system, shown in Fig.2, consists of two area 2
fully symmetrical areas linked together by two 230 kV lines of
220 km length. Each area is equipped with two identical round
rotor generators rated 20kV/900MVA. The synchronous TABLE III.
machines have identical parameters, except for inertia which PSS DAMPING ON A TWO TIE LINES SYSTEM
is H = 6.5s for generators in area 1 and H = 6.175s for Mode Open Loop FLPSS IEEE PSS4B IEEE PSS2B
generators in area 2. Thermal plants having identical speed Hz ζ Hz ζ Hz ζ Hz ζ
regulators are further assumed at all locations, in addition to Inter- 0.64 -0.026 0.55 0.19 0.52 0.31 0.62 0.1
fast static exciters with a 200 gain. The reference load-flow area
with generator G2 considered as the swing bus is such that all Local 1.13 0.096 0.83 0.6 1.92 0.47 1.06 0.32
area 1
generators are producing about 700 MW each. Two stressed Local 1.16 0.092 0.86 0.65 1.99 0.45 1.07 0.33
scenarios were considered with respectively, two tie-lines at a area 2
413 MW transfer level (K2L) and a single tie-line at a 353
MW transfer level (K1L). The loads are assumed everywhere For the single tie line system K1L inter-area frequency is
as constant impedance load models. The area 1 and area 2 0.44Hz without PSS and for the two tie lines system K2L
loads are 967 MW and 1767 MW respectively. The load inter-area frequency is 0.64 without PSS. The closed loop
voltage profile was improved by installing 187 Mvar more inter-area frequency is quite different for FLPSS and PSS4B
capacitors in each area. (Table II and II).
The damping performance of the FLPSS is compared in
Table II and Table III against the IEEE PSS4B and IEEE
PSS2B for these two systems. The IEEE4B PSS is clearly the
best for the inter-area modes. However, for the two local
modes the better damping is obtained with FLPSS. This good
performances hold at both week (K1L) and strong (K2L)
systems.
To illustrate the basic differences between these
stabilizers, a frequency response plot is shown on Fig.3 to
compare FLPSS settings with IEEE stabilizers (PSS4B and
PSS2B type).
Fig. 2. Four Machine Two-Areas Kundur test system
 4

Test A results illustrating the FLPSS performances are

shown in Fig.4. While all candidate devices perform well on a
two tie-line system, the FLPSS outperform the PSS4B and
PSS2B.

Fig. 3. Frequency response of the three PSS type

As shown in Fig.3, in the interest frequency range [0.1 Hz

4 Hz], the FLPSS, IEEE PSS2B and the IEEE PSS4B don't
present phase delay. Otherwise, it appears that, compared to
the other stabilisers, the FLPSS gain varies little between 0.04
Hz and 0.3 Hz. In this frequency interval the gain of the
FLPSS is located around 15p.u. However, beyond 1Hz, the
gain of the FLPSS increases quickly to reach 75p.u. around 4
Hz. To this frequency, the other stabilisers gains are 55p.u. Fig. 4. Test A: Inter-area angle shift, speed deviation and G1 terminal voltage of
(PSS4B) and 60p.u.(PSS2B). the K2L system

As a general observation, it is interesting to mention that Test B results are shown in Fig.5. We can see that when
the IEEE PSS4B frequency response is well balanced at both the K1L test system is in closed-loop with the PSS2B, it lost
ends of the spectrum, limiting the gain on the high side and the stability. The two other stabilizers (FLPSS and PSS4B)
phase lead on the low side. On the other hand, IEEE PSS2B succeeded in consolidating the network. For this contingency,
phase in the interval [0.01Hz 0.04Hz] is too large. we can say that the FLPSS and the PSS4B have broadly
comparable performances.
B. Large signal analysis
We have implemented the FLPSS described in previous
sections on G1, G2, G3 and G4 (as in the small signal
analysis), in order to analyze its behavior following large
contingencies (Table IV).
Two scenarios were considered respectively, two tie-lines
system (K2L) with a 413 MW transfer power and a single tie-
line system (K1L) with a 353 MW transfer power. The
contingencies applied to Kundur test system are presented in
table IV.
TABLE IV.
LARGE SIGNAL TESTS ON THE KUNDUR TEST SYSTEM

TEST SYSTEM CONTINGENCY DESCRIPTION

A K2L 15 cycles 3φ short-circuit at bus 8 cleared

system with one tie-line outage (line 7-9);
B K1L 12 cycles 3φ short-circuit at bus 1 (near
system G1) normally cleared with no equipment
outage;
C K2L 9 cycles 3φ short-circuit at bus 1 (near G1)
system normally cleared with no equipment
outage;
Fig. 5: Test B: Inter-area angle shift, speed deviation and terminal voltage (G1)
The FLPSS performances are exposed by comparing of the K1L system
analysis with reference PSS: PSS2B and PSS4B. Test C results are shown in Fig. 6. The K2L test system is
A severe fault was applied at the middle of the tie lines, less oscillating with the PSS4B and the FLPSS. We also note
followed by one tie-line outage (test A, Table IV). that the speed excursions are less pronounced with the
FLPSS, allowing the steady state to be reached more quickly.
 5

To complete this study, Test E results are shown in Fig.8.

In the case of a mechanical power negative ramp applied to
the generator G1, the PSS4B has better performances than the
other two stabilizers. On this type of event, the least powerful
stabilizer is the PSS2B because its stability signal is subject to
very low frequency oscillations, a probable side-effect of its
excessive phase lead below 0.04Hz.

Fig. 6. Test C: Inter-area angle shift, speed deviation and G1 terminal voltage of
the K2L system

C. Others important tests

Two other interesting contingencies for assessing PSS
behavior under mechanical power disturbances are described
Table V. In particular, Test D (Table V) results are shown in
Fig.7. The initial crest of the speed obtained with the FLPSS,
is smaller than with the two other stabilizers. Meanwhile, the
terminal voltage is not degraded too much which is a welcome
property. We also note that the FLPSS permits to reach the Fig. 8: Test E: PSS output signal, speed deviation and G1 Electrical power of the
K2L system
steady state faster. This said, Fig.7 also shows that the PSS2B
and PSS4B PSS have an equally acceptable performance for
IV. CONCLUSIONS
this Test D.
In this paper, fuzzy logic was used to synthesize a power
TABLE V. system stabilizer in order to maintain the stability of the
OTHERS TESTS ON THE KUNDUR SYSTEM power system over a wide operating range. The results
obtained in simulations are promising. Indeed, they showed
TEST SYSTEM CONTINGENCY DESCRIPTION
that the Fuzzy Logic based PSS satisfies two essential
D. K1L system Finite impulse (10% from reference properties in the control system field: good robustness and
mechanical power (Pref) for 1 second) applied good damping performance. It is important to emphasize that
on a single unit (G4) in this study no robust control technique was explicitly used.
E K2L system Mechanical power negative ramp
(1p.u./minute for 6 seconds) on a single unit
It demonstrates the potential and efficiency of fuzzy logic in
(G1) the power grid control field. Two modern PSS, the IEEE
PSS2B and IEEE PSS4B have been used in order to validate
the new FLPSS concept.

V. APPENDIX

A. IEEE PSS4B
TABLE VI. PSS4B PARAMETERS
Low Intermediate High
Frequency Frequency Frequency
Band Band Band
KB=20.0 KI=40.0 KH=160.0
KL1=66.0 KI1=66.0 KH1=66.0
KL2=66.0 KI2=66.0 KH2=66.0
TL1=0.4843 TI1=0.0969 TH1=0.0101
TL2=0.5812 TI2=0.1162 TH2=0.0121
TL7=0.5812 TI7=0.1162 TH7=0.0121
TL8=0.6974 TI8=0.1395 TH8=0.0145
VLmax=0.6 VImax=0.6 VHmax=0.6
Fig. 7: Test D: PSS output signal, speed deviation and G4 terminal voltage of the VLmin=-0.6 VImin=-0.6 VHmin=-0.6
K1L system KL11=1; KI11=1; KH11=1; VST=±0.15
 6

VI. REFERENCES
[1] R. Grondin, I. Kamwa, G. Trudel, L. Gérin-Lajoie, J. Taborda, “Modeling
and Closed-Loop Validation of a New PSS Concept, The Multi-Band
PSS,” Presented at the 2003 IEEE/PES General Meeting, Panel Session
on New PSS Technologies, Toronto, Canada
[2] IEEE Standard 421.5, IEEE Recommended Practice for Excitation
Systems Models for Power System Stability Studies, August 1992.
[3] M. Klein, G.J. Rogers, S. Moorty, P. Kundur, “Analytical Investigation of
Factors Influencing Power System Stabilizers Performance,” IEEE Trans.
on Energy Conv., 7(3), Sept. 1992, pp.382-390.
[4] P. Kundur, Power System Stability and Control, McGraw-Hill, New
York, NY, 1994.
[5] C. C.Lee, Fuzzy Logic in Control Systems : Fuzzy Logic Controller, Part I
and II, IEEE Transactions on Systems, Man, and Cybernetics, vol. 20, no.
2, pp. 404-435, 1990.
[6] H. X. Li, H. B. Gatland, “Enhanced Methods of Fuzzy Logic Control, “
IEEE transactions on systems, man and cybernetics, pp.331-336, 1995.
[7] N. Martins, A.A. Barbosa, J.C.R. Ferraz, M.G. dos Santos, A.L.B.
Bergamo, C.S. Yung, V.R. Oliveira, N.J.P. Macedo, "Retuning Stabilizers
for the North-South Brazilian Interconnection, " IEEE PES Summer
Meeting, 18-22 July 1999 , Vol. 1, pp. 58–67.
[8] MATLAB SimPowerSystem Software (version 6.5.1, 2004), MathWorks,
Fig.9: PSS4B
www.mathworks.com/products/simpower/
B. IEEE PSS2B [9] I. Kamwa, L. Gérin-Lajoie, "State-Space Identification-Towards MIMO
Models for Modal Analysis and Optimization of Bulk Power Systems,"
IEEE Trans. on Power Systems, 15(1), Feb. 2000, pp. 326-335.

VII. BIOGRAPHIES
Manuela Dobrescu received a B.Eng. (1989) in Electrical Engineering from
Craiova University, Romania and a M.Sc.(2003) from École Polytechnique,
Montreal University, Canada. In 1989, she joined RENEL-Romania, where she
was involved in the control and protection field for power plant generators and
high voltage stations. Twelve years later, she joined Hydro-Québec Research
Institute, where she is now involved as a Researcher in the field of power system
Tw ⋅ s T1 ⋅ s + 1 T3 ⋅ s + 1 T5 ⋅ s + 1 dynamics.
PSS = Ks1 ⋅ ⋅ ⋅
Tw ⋅ s + 1 T2 ⋅ s + 1 T4 ⋅ s + 1 T6 ⋅ s + 1 Innocent Kamwa (S'83, M'88, SM’98) received a PhD in electrical
engineering from Laval University, Québec, Canada, 1988, after graduating in
Fig.10: PSS2B 1984 at the same university. Since then, he has been with the Hydro-Québec
Research Institute, where he is at present a Principal Researcher with interests
TABLE VII. broadly in bulk system dynamic performance. Since 1990, he has held an
PSS2B PARAMETERS associate professor position in Electrical Engineering at Laval University where
five students have completed their PhD under his supervision. A member of
PSS2B Function PSS CIGRÉ, Dr. Kamwa is a recipient of the 1998 and 2003 IEEE PES Prize Paper
parameters parameters Awards and is currently serving on the System Dynamic Performance Committee
AdCom.
Tw1(2,3) 10.0 Ks1 20.0
Tw4 100 Tw 3
T8 0.50 T1 0.12
T9 0.10 T2 0.012
T3 0.12
M 5
T4 0.012
N 1
T5 0.25
Ks2 0.99 T6 0.75
Ks3 1.0 VST ±0.15

View publication stats