Está en la página 1de 9

Electrical Power and Energy Systems 46 (2013) 175183

Contents lists available at SciVerse ScienceDirect

Electrical Power and Energy Systems


journal homepage: www.elsevier.com/locate/ijepes

Effects of various power system stabilizers on improving power system


dynamic performance
Ping He a,b, Fushuan Wen c,d,, Gerard Ledwich c, Yusheng Xue e, Kewen Wang f
a

School of Electrical Engineering, South China University of Technology, Guangzhou 510640, China
College of Electrical and Information Engineering, Zhengzhou University of Light Industry, Zhengzhou 450002, China
c
School of Electrical Engineering and Computer Science, Queensland University of Technology, Brisbane, Queensland 4001, Australia
d
School of Electrical Engineering, Zhejiang University, Hangzhou 310027, China
e
State Grid Electric Power Research Institute, Nanjing 210003, China
f
School of Electrical Engineering, Zhengzhou University, Zhengzhou 450001, China
b

a r t i c l e

i n f o

Article history:
Received 14 June 2012
Received in revised form 27 September
2012
Accepted 9 October 2012
Available online 23 November 2012
Keywords:
Power system stabilizer
Eigenvalue analysis
Small/large-signal stability
Electromechanical modes

a b s t r a c t
To ensure the small-signal stability of a power system, power system stabilizers (PSSs) are extensively
applied for damping low frequency power oscillations through modulating the excitation supplied to
synchronous machines, and increasing interest has been focused on developing different PSS schemes
to tackle the threat of damping oscillations to power system stability. This paper examines four different
PSS models and investigates their performances on damping power system dynamics using both smallsignal eigenvalue analysis and large-signal dynamic simulations. The four kinds of PSSs examined include
the Conventional PSS (CPSS), Single Neuron based PSS (SNPSS), Adaptive PSS (APSS) and Multi-band PSS
(MBPSS). A steep descent parameter optimization algorithm is employed to seek the optimal PSS design
parameters. To evaluate the effects of these PSSs on improving power system dynamic behaviors, case
studies are carried out on an 8-unit 24-bus power system through both small-signal eigenvalue analysis
and large-signal time-domain simulations.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
Power systems worldwide have been continuously expanding
in scale and evolving into more complicated structures in past decades, and have to operate more frequently close to their limits as
the results of geographical and physical limitations as well as the
power industry restructuring. The secure operation of power systems, therefore, has become a major concern, and the applications
of power system stabilizers (PSSs) for dynamic stability enhancement have drawn more attention than ever before [110]. Conventional lead-lag PSSs (CPSSs) have been widely used by electric
utilities for this purpose. A PSS is used to provide some supplemental damping to rotor oscillations via an electric torque which is in
phase with the speed deviation [1]. In view of the fact that power
systems are highly nonlinear and operating conditions can vary
over a wide range, various kinds of PSSs have been developed in
the past decades, such as the xed parameter decentralized PSS,
the adaptive PSS, and the fuzzy logic based PSS, to name a few
[26].

Corresponding author at: School of Electrical Engineering, Zhejiang University,


Hangzhou 310027, China. Mobile: +86 13968105384; fax: +86 571 87952014.
E-mail address: fushuan.wen@gmail.com (F. Wen).
0142-0615/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijepes.2012.10.026

In order to take into account more system operating conditions,


the probabilistic eigenvalue analysis method was proposed for
designing power system damping controllers [1113]. With this
approach, the system stability is reinforced by shifting the distribution ranges of the critical eigenvalues toward the left side of the
complex plane. Coordination of the controller parameters was
achieved through solving a non-linear programming problem, in
which the objective function is composed of all unsatisfactory
eigenvalues. The objective function is minimized by using optimization approaches such as the steepest descent (SD) method so
that the overall performance of the controller could be optimized
under the given system states.
Up to now, various PSS design methods have been proposed and
some applied to different degrees in actual power systems [3
8,11,12]. CPSSs are generally based on xed parameters, and it is
hence not yet possible to achieve the optimal behavior for various
operating conditions of a power system. The adaptive power system stabilizers, as reported in [36], could track the changes of system dynamics in real time, and hence could perform well for
various operating conditions in principle. In 2003, a novel PSS
architecture was proposed in [7] and later included in the revised
IEEE Std-421.5 as PSS4B in [15]. Up to now, several kinds of PSSs
have been proposed, and it is not clear about their relative performances. With the development of large-scale power systems, a

176

P. He et al. / Electrical Power and Energy Systems 46 (2013) 175183

comparative analysis of some major kinds of PSSs representing the


key equipment for mitigating low frequency oscillations is very
demanding. Given this background, four major PSS design approaches are examined in this work. They are the CPSS, the Single
Neuron Model based PSS (SNPSS) [3], the adaptive PSS (APSS) [4,6],
and the multi-band PSS (MBPSS) [7,8]. The major consideration of
selecting these four kinds of PSSs for comparisons lies in that CPSS
and MBPSS are used most widely, while SNPSS and APSS are receiving extensive concerns. The performance of each PSS model is compared against that of others with regards to efciency and
robustness. An 8-unit 24-bus power system is used to perform
the small-signal eigenvalue analysis and large-signal time domain
simulations on each of the four PSS models. To make the comparisons equitable, the parameters of the four kinds of PSSs are all
determined by the steepest descent method [11].
This paper is organized as follows. Section 2 briey explains the
probabilistic eigenvalue analysis and optimization method. The
models of the four kinds of PSSs are presented in Section 3 and
the small-signal eigenvalue analysis is carried out in Section 4. In
Section 5, the large-signal time domain simulation is done, and
simulation results obtained. The paper is concluded in Section 6.
2. System model and optimization
The power system dynamic model can be linearized around the
system operating point, thus the state space equation set of a
power system can be expressed with the state variable vector X
and the system matrix A as shown below.

n0k nk  4rnk P nc


n nk  nc =rn P 4
k

where nc is the acceptable threshold of the damping ratio; in order


to ensure the system dynamic performance, nc should not be less
than a specied value (in this study, nc is specied to be 0.1) [14].
2.2. The optimization method
The sensitivities of a0k and n0k against the PSS gain reect which
generators PSS is more effective to stabilize the power system concerned, the probabilistic sensitivity indexes (PSIs) which are
a0

n0

Skmk @ a0k =@K m and Skkm @n0k =@K m (K m is a parameter of the mth
PSS) presented in [12] are thus used to determine the PSS siting
and the parameters under probabilistic conditions.
For the parameter tuning problem, an optimization model from
[11] as shown in Eq. (5), is employed here, in which only the
weak eigenvalues (ak < 4 or nk < 4) are included

Minimize FK

ak  42

ak <4

nk  42

nk <4

where FK is the objective function, K denotes the PSS parameter


vector whose practical limits depend on the hardware restrictions,
ak and nk are the functions of K. The optimization model represented by Eq. (5) can be solved by the steepest descent (SD) approach. Starting from an initial point K(0), the iteration proceeds
as below
i

X_ AX

By considering multiple operating conditions of a power system, all nodal injection power can be regarded as random variables
with the statistical attributions being determined by the probabilistic load ow. The corresponding eigenvalues can also be random
variables, even though their distribution characteristics are different. The probabilistic distributions and the stability probabilities of
all the eigenvalues can be obtained by means of the probabilistic
eigenvalue analysis [14].

K i1 K i  l1 rFK i

6
(i)

where i denotes the iteration number, rF(K ) is the gradient of


i
i
F(K) at point K(i), i.e. rFK i @FK=@KjKK , and l1 is the optimal step size along the direction rF(K(i)) and is obtained by running the one-dimensional search algorithm [12].
3. PSS moldeling
This section presents the models of the four PSSs to be
evaluated.

2.1. Probabilistic eigenvalue analysis


3.1. The conventional PSS (CPSS)
The statistical characteristics of a random variable can be
described by its corresponding numerical characteristics such as
the mathematical expectation (mean value), variance and covariance. Similarly, the statistical characteristics of an eigenvalue can
be represented similarly by these characteristics [14].
Under the assumption of the normal distribution, for a particular eigenvalue kk ak jbk ; under the 4r criterion the real part
 k and the standard deviation rak will distribwith the expectation a
 k  4rak ; a
 k 4rak g with the probability 0.99993). To
ute within fa
 k  4rak ; a
 k 4rak g should be located
ensure the stability of kk ; fa
on the left side of the complex plane, which can also be described
by the upper limit a0k and the standardized real part expectation ak
as shown below [11].

a0k a k 4rak 6 0
ak a k =rak P 4

Therefore, a0k and ak can be regarded as two extended damping


coefcients from which the robust stability of kk can be estimated.
The damping ratio for an eigenvalue kk is dened as

q
nk ak = a2k b2k

Similar to the calculation of a0k and ak in Eq. (2), the expected
damping ratios can be calculated with the expectation 
nk and the
standard deviation rnk as

For simplicity, the CPSS is modeled by two identical lead/lag


networks represented by a gain KPSS, four time constants Tl, T2, T3
and T4, and a washout circuit taking a time constant Tw [15]. The
transfer function of CPSS, G(s), can be expressed as Eq. (7).

Gs K PSS

sT w 1 sT 1 1 sT 3
1 sT w 1 sT 2 1 sT 4

3.2. The single neuron based PSS (SNPSS)


Due to its strong ability to self-study and self-adapt, a single
neuron [16] has been applied in power system controllers for
achieving a better performance. With its simple algorithm and
working principle, the single neuron has the structure as illustrated
in Fig. 1.
In Fig. 1, yt is the system output; rt is the desired or reference value; kt stands for the gain; xi t stands for the ith input
signal; wi t is the connection weight corresponding to xi t; A is
the converting device; B is the controlled object, and stand for a
power system with PSSs installed.
The learning algorithm of the single neuron can be found in [3],
and will not be detailed in this paper.
The control signal of a PSS, such as the generator speed deviation Dx, or the power deviation DP, can be used as the input signal

P. He et al. / Electrical Power and Energy Systems 46 (2013) 175183

177

and KN are the gains of different frequency subbands; TA, TB and TN


are the corresponding time constants; Tw is the washout time
constant.
4. Small-signal eigenvalue analysis
This section performs small-signal stability analysis for each of
the four kinds of PSSs.
4.1. The test system

Fig. 1. The conguration of the single neuron control system.

Fig. 2. The structure of the SNPSS.

of the signal neuron. In this study, DP is employed, and DP is zero


under the normal operating conditions. The input signal x1 t; x2 t
and x3 t of the converting device A can be represented as

8
>
< x1 t rt 0
x2 t rt  yt DP
>
:
x3 t x2 t  x2 t  1 Dx2 t @ DP=@t

Under small disturbances, the equation set describing the system dynamic states can be linearized around a given operating
point. An algorithm such as Eq. (8) can be represented by the form
of a rst-order transfer function block. Together with the transfer
function G(s) of a PSS, the combined model is shown in Fig. 2. An
auxiliary PSS gain kt can be obtained by multiplying the gain of
the single neuron and G(s). For the convenience of presentation,
in this work, kt is considered the function of the generator active
power Pt, which will be self-adjusted with the changes of the systems operating conditions.
3.3. The adaptive PSS (APSS)
Fig. 3 shows the conguration of APSS [46] using three design
points. The weighing-coefcients W1, W2 and W3 are employed to
simulate the actual operating conditions.
3.4. The multi-band PSS (MBPSS)

An 8-unit 24-bus power system as shown in Fig. 5 is used as the


study system [13]. Each generator is represented by a sixth-order
model, and the models of the excitation system as well as the turbine governor are given in the Appendix. Table 1 lists the electromechanical modes for the situation without PSS installed (NPSS).
720 samples are taken from the standardized daily operating
curves of this system to simulate multiple system operating conditions (Fig. 6).
The probabilistic sensitivity indices as described in Section 2 are
listed in Table 2. Six PSSs are installed on G1, G2, G3, G5, G6 and G7 of
this 8-unit system. The results in Table 2 shows that PSSs at G3, G5,
G6 and G7, especially G7, are more sensitive in the electromechanical modes. Hence, the four kinds of PSSs are separately applied to
PSS7 to test their effectiveness. The PSS parameters obtained by the
preceding probabilistic method in [11] are used as the initial values
of the succeeding method, and optimized by the steepest descent
approach as described in Section 2.2.
Table 3 lists the results of seven electromechanical modes with
the CPSS.
4.2. SNPSS
As mentioned before, 720 samples are taken from the standardized daily operating curves of Fig. 6. In order to identify the auxiliary PSS gain kt, the PSS parameters are optimized under 720
operating conditions, where each sample of kt is the optimum value under each of the 720 system conditions by minimizing the
objective function of Eq. (5). 720 PSS7 samples are then obtained,
including the auxiliary gain k7(t) and the corresponding generators
active power. To better verify the self-regulation of the auxiliary
PSS gain, P3(t) (the power generation of G3) is selected because of
its relatively large uctuation. From the 720 samples, the relationship between k7(t) and P3(t) can be obtained by curve tting using
the least squares method. By numerous simulation comparisons,
the four-order tting function produces the least error compared
to the linear, second-order, cubic and exponential tting functions.
The four-order tting function thus obtained is

k7 t 15222:9921 14203:580  P 3 t  4950:9448  P3 t2


The main characteristics of the MBPSS model [7,8] examined in
this work are shown in Fig. 4. K is the main gain of the PSS; KA, KB

764:59  P3 t3  44:1415  P 3 t4

Fig. 3. The APSS block diagram.

178

P. He et al. / Electrical Power and Energy Systems 46 (2013) 175183

L2
L7

Power (p.u.)

2.5
2
1.5
1
0.5
0

10

12

14

16

18

20

22

24

time (h)
Fig. 4. The MBPSS block diagram.

Fig. 6. The standard daily operating curves.

Table 2
Probabilistic sensitivity indices of the critical damping ratios to PSS gains.

n1
n2
n3
n4
n5
n6
n7

K1

K2

K3

K4

K5

K6

K7

.1378
.0007
.0003
.0000
.0006
.0004
.0244

.0005
.1235
.0031
.0000
.0001
.0003
.0057

.0005
.0017
.0024
.0017
.3370
.0094
.2473

.0001
.0003
.0000
.0636
.0164
.0331
.0462

.0002
.0005
.0001
.0762
.0107
.0701
.0959

.0004
.0422
.2262
.0000
.0007
.0010
.9789

.0005
.0012
.0033
.0002
.0021
.1046
.2549

By applying Eq. (10) in PSS7, seven electromechanical modes


and the eigenvalues with ak < 4 or nk < 4 are obtained, as listed
in Table 5.

4.4. MBPSS
Fig. 5. The 8-unit 24-bus power system.

Applying Eq. (9) in PSS7 and performing the probabilistic eigenvalue computation, seven electromechanical modes are obtained
as given in Table 4.
4.3. APSS
Three typical time points, i.e. 0 am, 9 am and 15 pm are employed, and the weighting coefcients in Fig. 3 are specied as
W1 = 0.3625, W2 = 0.4142 and W3 = 0.2234 [4]. The output of the
APSS can be expressed as

GPSS7 0:987

5s 1 1:153s 1 0:052s
1 5s 1 0:9s 1 0:05s

10

In Fig. 4, the washout time constant Tw usually takes 310 s. In


this work, Tw is specied to be 5 s, and the other time constants can
be calculated by Eq. (11).

8
p
p
>
< T A2 T A3 1=2pfA RA 1=xA RA
T A1 TRA2A
>
:
T A4 T A2 RA

11

where xA is the A-band central angular frequency, and RA contains


the compensation coefcients of the A-band [7,8]. In this work, PSS7
is designed with the two-band structure, with xM 4:5; xH 50;
K M 5; K H 25; RM 1:18 and RH 1:14. The seven electromechanical modes are listed in Table 6.

Table 1
Electromechanical modes without PSS (NPSS).
Mode

a


b

ra

Pa


n

rn

Pn

1
2
3
4
5
6
7

1.731
0.757
0.565
0.637
0.650
0.421
0.010

15.854
11.113
9.840
7.908
7.530
6.482
3.823

0.0439
0.0827
0.0491
0.0172
0.0985
0.0306
0.0117

38.33
8.55
10.48
34.18
6.09
12.11
3.43

1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
0.0003

0.1085
0.0680
0.0573
0.0803
0.0860
0.0647
0.0026

0.0033
0.0072
0.0037
0.0036
0.0112
0.0047
0.0030

2.57
4.46
11.44
5.55
1.26
7.50
31.96

0.9949
0.0000
0.0000
0.0000
0.1057
0.0000
0.0000

 
a ; b;
n are the expectations of the real part, the imaginary part and the damping ratio of eigenvalue; ra ; rn are the standard deviation of k; a ; n are the standardized
 =ra ; n 
expectation of k; a ac  a
n  nc =rn , and aC 0:01; nC 0:1; P a Pa < aC and P n Pn > nC are the stability probabilities of k.

179

P. He et al. / Electrical Power and Energy Systems 46 (2013) 175183


Table 3
Electromechanical modes with CPSS installed.
Mode

a


b

ra

a

Pa


n

rn

n

Pn

1
2
3
4
5
6
7

2.417
1.434
1.213
1.558
0.838
1.766
1.013

15.030
10.672
9.682
9.122
7.743
7.334
4.198

0.0473
0.0336
0.0472
0.1337
0.0145
0.1755
0.1588

50.04
41.20
24.64
11.27
54.23
9.78
6.06

1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000

0.1587
0.1331
0.1243
0.1683
0.1076
0.2341
0.2345

0.0048
0.0052
0.0028
0.0102
0.0009
0.0180
0.0334

12.15
6.34
8.69
6.67
8.95
7.44
4.02

1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000

Table 4
Electromechanical modes with SNPSS installed.
Mode

a


b

ra

a

Pa


n

rn

n

Pn

1
2
3
4
5
6
7

2.418
1.445
1.211
2.174
1.750
0.873
1.276

15.030
10.670
9.671
9.314
8.152
7.759
4.586

0.0464
0.0257
0.0466
0.2005
0.0887
0.0115
0.1028

51.08
54.19
24.92
10.59
19.16
71.38
11.93

1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000

0.1589
0.1342
0.1242
0.2273
0.2099
0.1119
0.2681

0.0047
0.0045
0.0029
0.0334
0.0199
0.0013
0.0274

12.40
7.67
8.38
3.81
5.52
9.10
6.14

1.0000
1.0000
1.0000
0.9999
1.0000
1.0000
1.0000

Table 5
Electromechanical modes with APSS installed.
Mode

a


b

ra

Pa


n

rn

n

Pn

1
2
3
4
5
6
7

2.420
1.453
2.325
1.208
1.520
0.889
1.138

15.031
10.667
9.745
9.667
8.206
7.765
4.669

0.0456
0.0222
0.2451
0.0414
0.1256
0.0095
0.0816

51.95
63.26
9.28
27.99
11.71
88.15
13.34

1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000

0.1589
0.1350
0.2321
0.1239
0.1822
0.1137
0.2369

0.0047
0.0039
0.0323
0.0026
0.0222
0.0015
0.0217

12.59
9.08
4.09
9.39
3.70
9.08
6.30

1.0000
1.0000
1.0000
1.0000
0.9999
1.0000
1.0000

Table 6
Electromechanical modes with MBPSS installed.
Mode

a


b

ra

a

Pa


n

rn

n

Pn

1
2
3
4
5
6
7

1.880
1.626
1.346
0.888
1.192
1.700
1.483

15.995
11.437
9.966
7.825
7.690
5.930
4.170

0.0547
0.0831
0.0947
0.0040
0.1101
0.1899
0.1175

34.38
19.57
14.22
219.39
10.82
8.95
12.63

1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000

0.1167
0.1408
0.1339
0.1127
0.1531
0.2756
0.3352

0.0041
0.0078
0.0080
0.0017
0.0117
0.0248
0.0227

4.04
5.19
4.26
7.48
4.52
7.08
10.36

1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000

4.5. Performance comparisons


By comparing Tables 36 with Table 1, it is found that the
damping ratios of most electromechanical modes are signicantly
enhanced.
Tables 3 and 4 give the probabilistic results of the seven electromechanical modes with the CPSS and the SNPSS, respectively. In
general, the SNPSS performs better except for a comparatively
large drop of 
n6 which decreases from 0.2341 to 0.1119. It should
be pointed out, nevertheless, that 
n6 0:1119 > 
nC 0:1 does satisfy the acceptable threshold of the damping ratio, and that more
importantly, 
n4 ; 
n5 and 
n7 have been signicantly enhanced. In
practice, it is not easy to receive a timely remote signal due to communication network delays and some other difculties. The APSS is
designed to adapt to these working conditions. Table 5 shows that
the damping ratios of modes 3, 5, and 7 are enhanced with the
APSS. Similarly, according to the results in Table 6, all electromechanical modes have sufcient stability margins, though some
are slightly reduced, such as 
n1 and 
n4 , but still satisfy the accept-

able threshold. More importantly, the damping ratio of mode 7 is


increased by 43% compared with that of the CPSS.
Assuming a normal distribution, the probabilistic density functions (PDFs) of the real part and the damping ratio in modes 5 and
7 with different PSSs are shown in Figs. 7 and 8. As revealed by
Fig. 7, the performances of mode 5 with SNPSS is better than that
of other PSSs. Fig. 8 illustrates that the MBPSS is more robust as
the corresponding eigenvalues fall into the region with larger stability margin.
Table 7 outlines the degrees of improvement of real parts and
damping ratios when using SNPSS, APSS and MBPSS over CPSS, in
which the negative values (shown in bold) indicate the degrees
of decline. It can be seen from the data of Table 7 that mode 6 is
deteriorated with SNPSS installed, modes 4 and 6 worsened with
APSS installed, and the damping ratios of modes 1 and 4 are reduced with MBPSS installed.
As shown in Table 2, the sensitivities of modes 6 and 7 against
the PSS7 are 0.1046 and 0.2549, respectively, and this means
that PSS7 inuences modes 6 and 7 more than the other modes.

180

P. He et al. / Electrical Power and Energy Systems 46 (2013) 175183

Fig. 7. The PDF curves of mode 5 with different PSSs installed on G7.

Fig. 8. The PDF curves of mode 7 with different PSSs installed on G7.

Table 7
The improvement degrees of real parts and damping ratios when using SNPSS, APSS and MBPSS over CPSS.
Mode

1
2
3
4
5
6
7

Real part improvement (%)

Damping ratio improvement (%)

SNPSS

APSS

MBPSS

SNPSS

APSS

MBPSS

0.04
0.77
0.16
39.5
109
50.1
25.9

0.13
1.32
91.7
22.5
81.3
49.6
12.3

22.0
13.4
11.0
43.0
42.2
3.74
46.4

0.13
0.83
0
35.1
95.1
52.2
14.3

0.13
1.43
86.7
26.4
69.3
51.4
1.02

26.5
5.79
7.73
33.1
42.3
17.7
42.9

Fig. 9. The relative rotor angle curves between G1 and G8 under a three-phase grounding short-circuit fault with different kinds of PSSs respectively installed.

P. He et al. / Electrical Power and Energy Systems 46 (2013) 175183

181

Fig. 10. The relative rotor angle curves between G6 and G8 under a three-phase grounding short-circuit fault with different kinds of PSSs respectively installed.

Fig. 11. The voltage on bus 1 under a three-phase grounding short-circuit fault with different kinds of PSSs respectively installed.

Fig. 12. The voltage on bus 11 under a three-phase grounding short-circuit fault with different kinds of PSSs respectively installed.

From Tables 3 and 7, it is known that the performance of mode 7


has more or less improved with four kinds of PSSs. However, only
MBPSS improves mode 6 relative to CPSS. It can be seen from Table
7 and Fig. 8 that the improvement of the damping characteristics
with MBPSS is better than that of other PSSs, namely SNPSS and
CPSS.
5. Large-signal stability assessment
All simulation results presented in this section are obtained
using the same power system as shown in Fig. 5. The parameters

of all generators as well as those of the various kinds of PSSs remain the same as in the situation of the small-signal stability analysis. A particular system operating condition from the 720 samples
is chosen to perform the large-signal stability time domain simulation, and bus 24 is specied to be the slack bus. A three-phase
grounding short-circuit fault occurred between buses 9 and 11
near bus 11 with the fault time tf = 1.0 s and clearing time
tc = 1.2 s. The behaviors of the four types of PSSs at G7 are simulated
in the time domain.
The relative rotor angle curves between G1, G6 and G8 with different PSSs respectively installed under a three-phase grounding

182

P. He et al. / Electrical Power and Energy Systems 46 (2013) 175183

Fig. 13. The active power curve of G3 under a three-phase grounding short-circuit fault with different kinds of PSSs respectively installed.

Fig. A1. The excitation system of generators G1, G2, G3, G6, G7 and G8.

Fig. A2. The excitation system of generators G4 and G5.

Fig. A3. The turbine governor of all generators.

short-circuit fault are given in Figs. 9 and 10, the voltages on buses
1 and 11 are shown in Figs. 11 and 12, and the active power curve
of G3 is shown in Fig. 13.
It can be observed from Figs. 913 that under the three-phase
grounding short-circuit fault, the relative rotor angle curves are
divergent in the case without PSSs installed, the bus voltage at
the short-circuit point fall down to zero, and nally the system
loses the stability at around 2 s. When the four kinds of PSSs are
respectively installed, the system tends to stabilize at around 5s.
From Figs. 9 and 10, it can be seen that the amplitudes of the power
angle curves with CPSS and APSS respectively installed are larger
than those with MBPSS and SNPSS respectively installed, and the
swing characteristics of the system with MBPSS and SNPSS respectively installed are similar. Similar situations are observed in the
prole of the active power curve of G3, as shown in Fig. 13. The

bus voltage curves with four kinds of PSSs respectively installed


exhibit similar features, as shown in Figs. 11 and 12.

6. Conclusions
This paper examined the effects of four different PSS design
approaches, namely CPSS, SNPSS, APSS and MBPSS, on improving
power system dynamic performances. The steepest descent method
is used to obtain the optimal PSS parameters. An 8-unit 24-bus sample power system is used to evaluate the system behaviors under
small and large disturbances with four types of PSSs respectively
installed. The eigenvalue analysis for the small disturbance scenario
shows that the damping ratios of most electromechanical modes
are signicantly enhanced with four kinds of PSSs respectively

P. He et al. / Electrical Power and Energy Systems 46 (2013) 175183

installed. Compared with CPSS, the performance of mode 7 with


MBPSS is more robust as the corresponding eigenvalues all fall into
the region with a larger stability margin, and the performance of
mode 5 with SNPSS is better than those of other PSSs. On the other
hand, it is shown by time-domain simulations for the large disturbance scenario that the amplitudes of the curves with CPSS and
APSS respectively installed are larger when a fault occurs, while
the swing characteristics with MBPSS and SNPSS respectively installed are similar. These results suggest that the performances of
MBPSS and SNPSS are superior to CPSS and APSS in improving the
amplitude oscillations of the system.
Acknowledgement
This work is supported by a discovery project of Australia Research Council (DP120101345).

Appendix A. Block diagram of the excitation systems and the


turbine governors
See Figs. A1A3.
References
[1] Kundur P, Klein M, Rogers GJ, Zywno MS. Application of power system
stabilizers for enhancement of overall system stability. IEEE Trans Power Syst
1989;4(2):61426.
[2] Nechadi E, Harmas MN, Hamzaoui A, Essounbouli N. A new robust adaptive
fuzzy sliding mode power system stabilizer. Int J Electr Power Energy Syst
2012;42(1):17.

183

[3] He P, Wang KW, Tse CT, Bian XY. Studies of the improvement of probabilistic
PSSs by using the single neuron model. Int J Electr Power Energy Syst
2007;29(3):21721.
[4] He P, Wei YB, Yang CX, Wang KW. Adaptive power system stabilizer design for
multi-machine power systems. In: Proceedings of 8th international conference
on advances in power system control, operation and management (APSCOM),
Hong Kong, China; November 811, 2009. p. 16.
[5] Chatterjee A, Ghoshal SP, Mukherjee V. Chaotic ant swarm optimization for
fuzzy-based tuning of power system stabilizer. Int J Electr Power Energy Syst
2011;33(3):65772.
[6] Chaturvedi DK, Malik OP. Neurofuzzy power system stabilizer. IEEE Trans
Energy Convers 2008;23(3):88794.
[7] Grondin R, Kamwa I, Trudel G, Gerin-Lajoie L, Taborda J. Modeling and closedloop validation of a new PSS concept, the multi-band PSS. In: Proceedings of
IEEE power engineering society general meeting, Toronto, (Canada); July 13
17, 2003. p. 18049.
[8] Kamwa I, Grondin R, Trude G. IEEE PSS2B versus PSS4B: the limits of
performance of modern power system stabilizers. IEEE Trans Power Syst
2005;20(2):90315.
[9] Gurrala G, Sen I. Power system stabilizers design for interconnected power
systems. IEEE Trans Power Syst 2010;25(2):104251.
[10] Radaideh SM, Mnejdaw I, Mushtaha MH. Design of power system stabilizers
using two level fuzzy and adaptive neuro-fuzzy inference systems. Int J Electr
Power Energy Syst 2012;35(1):4756.
[11] Tse CT, Wang KW, Chung CY, Tsang KM. Parameter optimization of robust
power system stabilizers by probabilistic approach. IEE Proc Gener Transm
Distrib 2000;147(2):6975.
[12] Chung CY, Wang KW, Tse CT, Bian XY, David AK. Probabilistic eigenvalue
sensitivity analysis and PSS design in multimachine systems. IEEE Trans Power
Syst 2003;18(4):143945.
[13] Wang KW, Chung CY, Tse CT, Tsang KM. Multimachine eigenvalue sensitivities
of power system parameters. IEEE Trans Power Syst 2000;15(2):7417.
[14] Wang KW, Tse CT, Tsang KM. Algorithm for power system dynamic stability
studies taking account the variation of load power. Electr Power Syst Res
1998;46(3):2217.
[15] IEEE recommended practice for excitation system models for power system
stability studies. IEEE Standard 421.5-2005.
[16] Mckenna T, Davis J, Zornetzer SE. Single neuron computation. San Diego
(CA): Academic Press; 1992.

También podría gustarte