A Comparative Analysis on Different Types of Power

System Stabilizers

Moudud Ahmed

Department of EEE

Bangabandhu Sheikh Mujibur Rahman Science and

Technology University

Gopalganj, Bangladesh

e-mail:[Link]@[Link]

Naruttam Kumar Roy

Department of EEE

Khulna University of Engineering and Technology

Khulna, Bangladesh

e-mail: nkroy@[Link]

Abstract—This paper presents a comparative analysis among

various Power System Stabilizers (PSSs) for the enhancement of

small signal stability of power systems. The most efficient approach

of countering small signal instability dilemma is to add a

supplementary controller known as PSS to the Automatic Voltage

Regulator (AVR) on the generator. The conventional PSS seems to

perform well at particular network configurations for which they

are tuned. However, when system parameters change due to

disturbance, unambiguous condition arises. The above limitations

can be overcome by improving PSS design technique using

artificial intelligence. In this paper, the effectiveness of various PSS

design

in MATLAB/Simulink

environment by performing nonlinear simulations utilizing a

synchronous machine-infinite bus model. Among various PSSs the

Adaptive Neuro Fuzzy Inference System (ANFIS) design technique

is found better compare to the other design techniques.

techniques has been analyzed

Keywords- Adaptive Neuro Fuzzy Inference System (ANFIS),

Fuzzy Logic Controller (FLC), Frequency Response, Power System

Stabilizer (PSS), PID controller, Root Locus .

I.

Nonlinear power

INTRODUCTION

systems

continuously

experience

generation and load change conditions along with severe

disturbances causing change in power system operating

conditions [1]. It is essential to keep voltage and frequency of

power systems in the desired level with a continuous balance

between electricity generation and load demand [2]. Due to the

enormous increase in demand for the electricity, the system

needs to operate close to the stability limit with quicker and

flexible manner in the deregulated competitive environment.

These requirements may cause instability as well as low

frequency electromechanical oscillations [3]. In this regard, fast

excitation control regulates the generated voltage and thus,

helps to control steady-state and transient stability of modern

synchronous generators [4]. In comparison to ammortisseur

winding and governor control, Automatic Voltage Regulators

(AVRs) are extremely suitable for regulation of voltage

through excitation control. Although AVRs are good to control

generated voltage, a high gain excitation system has an adverse

effect on

the dynamic stability of power system as

electromechanical oscillations of low frequencies continue for

long time and retard the capabilities of power transfer [5], [6].

In order to improve power system stability, a supplementary

stabilizing signal using Power System Stabilizer (PSS) has

been used in AVR to suppress oscillations of low frequency as

well as to improve damping of a synchronous generator [7].

Many techniques and approaches have been put to provide

the damping torque required for improving the power system’s

stability such as root locus design method, frequency response

lag-lead compensator which are known as

techniques,

conventional PSS (CPSS). The above mentioned CPSSs can be

designed for Heffron-Phillip’s model of single machine infinite

bus (SMIB) system. The CPSS parameters are derived from

iterative and needs huge

is

eigenvalue analysis which

computations [7], [8]. Furthermore, because of nonlinear

behavior of power system’s configurations, parameters are

changed from time to time [9]. Therefore, they are not able to

give optimal solution over wider ranges of operating

circumstances. A Fuzzy Logic PSS (FLPSS) has been

employed to improve the performance of power systems [6].

The advantage of fuzzy logic is there is no need of

mathematical model of the system [10]. However, in FLPSS

membership functions (MFs) are set according to trial and error

method. On the other hand, during modeling, it cannot be

predicted what the MF’s should look like simply from looking

at data [10]. In this regard, the requirement of an Adaptive

Neuro-Fuzzy Inference System (ANFIS) is obvious [11]. An

ANFIS based PSS is a new and promising method in which the

MF’s and rules are adjusted adaptively on the basis of input-

output data given as training data to ANFIS [12]. This method

is a combination of fuzzy logic with neural networks [13]. As

there are many techniques of PSS design, the comparative

analysis among different types of PSS is important to select a

suitable design technique for improving the performance of

power systems. The main contributions of this paper are to (a)

design different types of PSS and (b) analyze the performance

of different types of PSS design techniques.

This paper is arranged in five sections. In section II, a

concise review of the SMIB model and AVR is discussed. In

Section

techniques are discussed.

Consequently, in Section IV, Fuzzy logic and ANFIS based

PSS are discussed. Section V shows simulation results and

Section VI brings a conclusion of this work.

III, CPSS design

978-1-5090-1269-5/16/$31.00 ©2016 IEEE2016 5th International Conference on

Informatics, Electronics and Vision (ICIEV)797

II. POWER SYSTEM MODEL

III. CPSS DESIGN METHOD

A. Synchronous Generator Model

A. Root Locus Design

In this paper, Heffron-Phillip’s model of SMIB system

shown in Fig. 1 is used [2]. Here, KS is synchronizing

coefficient, KD is damping torque coefficient, H is inertia

constant, ∆wr is deviation of speed, wo is rated speed, s is

laplace operator, Gex is representing the transfer function of

AVR and exciter and K1 to K6 are known as K constants, KA is

exciter gain, TR, T3, and TA are time constant of voltage

transducer, field circuit, and exciter, respectively. An AVR

introduces synchronizing torque and damping torque which

solely depends on constant K5.. The numerical values of the

constants are given in Appendix.

Figure 1. Block diagram of SMIB system [2].

B. Proportional-Integral-Derivative (PID) Controller based

AVR

An AVR keeps constant terminal voltage as well as

regulates real power flow. An AVR with PID controller

improves dynamic response and reduces steady-state error. It

improves transient response by padding finite zeros to the plant

open loop transfer function (TF) [14]. It increases system type

by adding a pole at the origin. The TF of PID controller is

GC(s) = KP + k I /s + kD s. A PID controller is added in the

forward path of AVR. For tuning of PID parameters firstly, the

value of KP is selected and KI, KD is adjusted to some value and

then simulation is performed to observe overshoot and settling

time. The tuning process is continued until a suitable response

is obtained. The values of Kp, Ki and Kd are 1, 0.25 and 0.28,

respectively. The block diagram of AVR with PID controller is

shown in Fig. 2.

+ Vt

PID

Exciter

Generator

Vref -

Controller

Voltage

sensor

Figure 2. AVR system with a PID controller.

 It involves the design of a Proportional Integral (PI)

controller as an AVR and PSS using root locus. At the

starting point of this method, the AVR loop is replaced by

a PI controller and controller parameters are being selected

on the basis of time rise (tr) and maximum overshoot (Mp)

which must be less than 0.50 sec and 10%, respectively

[15], [16]. Then the step response of the regulator loop is

drawn from which steady-state error is examined. Figure 3

shows the step response for 0.1pu step input and different

values of Ki. For Kp=35 and Ki=0.6 the required condition

is satisfied. From the above plot, it is clear that there are

steady state error and oscillation due to swing mode.

Hence, to reduce oscillation a PSS is introduced in the

feedback path. Therefore, a PSS has to be designed by root

locus technique. The root locus of close loop AVR and

PSS is drawn and analyzed to find out the complex

dominant pole. It is found that the dominant complex poles

are at -0.4801±9.332i. From

the above pole using

MATLAB compute departure angle (Φp) from the pole

and it is Φp =43.28° [15]. According to the value of

departure angle the order and parameters of phase lead

compensator is chosen. Here, it is needed to add an angle

of 137º which cannot be done using a single lead

compensator. Therefore,

lead
compensators in series each adding an angle of 68.5º,

in

having

form P(s) = K

], such that

[Kα

Φp=180º [15]. Finally, the root locus and response of

SMIB system using this PSS is redrawn and examined.

Figure 4 shows the uncompensated system root locus plot

having dominant complex pole -0.4801±9.332i and Fig. 5

shows the final compensated system root locus plot [15].

function

][ Kα

transfer

+
z

/()

this paper uses

the

/()

two

r
p

0.12

0.10

0.08

0.06

0.04

0.02

0.00
0

ki=1

Ki=0.6

ki=0.4

ki=0.2

Ki=0.1

10

Time(sec)

Figure 3. Step response for 0.1pu step input and different values of Ki.

798
s

+
.0

/()

/()1.0

0125

this stage, the new gain crossover frequency is determined. To

provide the required phase lag to the system at the crossover

frequency, a lag-compensator is used as the AVR, having

transfer function V(s) = Kl

, where Kc is

required dc gain, compensator gain Kl=Kc/β, and p=z/β. The

lag required at the crossover frequency is 20 (log1/β) [15]. The

(35

. Now

resultant phase lag AVR is

step response is determined by adding the AVR loop. Figure 7

shows the step response of lag compensator as AVR for 0.1pu

step input where tr= 0.4810 sec and Mp= 6.7941% are obtained.

Therefore, the state-space model is generated from Vref to ω

(speed) with the regulation loop closed and the model is

decomposed to isolate the Q(s) path from the speed to electrical

torque (effect of speed on electric torque due to machine

dynamics) to obtain new matrix from the original matrix A. In

the next step, resulting state space model is converted to

transfer function and Q(s) is connected with torsional filter and

washout filter to get F(s). The frequency response plot of F(s)

is shown in Fig. 8 which shows phase at 2 rad/sec is -37º and

phase at 20 rad/sec is -105º. Therefore, a phase lead controller

has to be designed which increases this phase at 2 to 20 rad/sec

from the above values to approximately 0º to -15º, such that the

feedback loop adds pure damping to the dominant poles. It

needs an additional phase of 35º at 2 rad/sec, 60º at 12 rad/sec,

and 100º at 20 rad/sec [15]. Here, maximum phase addition

Φm is 100º at 20 rad/sec which is too large for a single lead

compensator. Therefore, this paper uses two compensator in

][

series, having transfer function V(s)= K[Kα

, z=1/T, p=1/αT, and Kα=1/α [15]. Finally,

response of the designed system is examined. The root locus

plot is shown in Fig. 9.

+

/()

−=m

1(

1/()

/()

α/1=

], where
sin

Figure 4. Uncompensated system root locus.

Figure 5. Final compensated system root locus plot.

B. Frequency Response Design

A frequency response design utilizes bode plot to measure

gain margin (Gm), phase margin (Pm), DC gain and

compensating phase by adding lag controller in the AVR and

lead controller in the PSS loop [16]. The frequency response

plot of uncompensated system in Fig. 6 shows the dc gain

(-2.59dB) 0.74 and magnitude peak at 9.3 rad/sec. A phase lag

AVR has to be designed provided that dc gain must be greater

than 200 and phase margin must be greater than 800 [16]. At
Figure 6. Frequency response plot with and without VR loop.

799

1
.

0.12

0.10

0.08

0.06

0.04

0.02

0.00

lag controller as AVR

4
6

10

Time(sec)

Figure 7. Step response of lag compensator as voltage regulator for 0.1pu step

input.

Figure 9. Root locus of PSS loop with compensator.

IV. PSS DESIGN USING ARTIFICIAL INTELLIGENCE

A. FLPSS Design

It is generally a rule base system that executes a control

decision on the basis of fuzzy rules [17]. A fuzzy inference

system of FLC maps the given input to fuzzy variables, fuzzy

variables to rule to produce consequent fuzzy region for each

variables, and finally, defuzzified outputs. The FLPSS input

signals are speed deviation of rotor and acceleration and

voltage stabilization signal to exciter is the output. Seven

triangular linguistic fuzzy subset for each of the input-output

variables are used [17]. The abbreviations of seven linguistic

variables are respectively NB (Negative Big), NM (Negative

medium), NS (Negative Small), ZE (Zero), PS (Positive

Small), PM (positive Medium) and PB (Positive Big). Two

inputs, speed deviation and acceleration generate 49 rules that

define the relation between input and output of fuzzy

controller. Thus, the stabilizing signals have been computed

[18]. The rule base of fuzzy logic controller is shown in Table

I. The design procedures of FLPSS are as follows:

Figure 8. Frequency response of damping loop.

 Select input and output variables.

 Select MFs.

 Set up fuzzy rule.

 Finally, choose defuzzification criteria.

800
TABLE I.

RULE TABLE OF FLC [18]

mentioned that CPSS parameters are tuned and works well for

a particular operating conditions. A FLPSS shows better

performance compared to CPSS, however, MF selection is a

tough task for it. From the comparison, it can be observed that

an ANFIS-PSS shows better performance in all respect. The

main advantage of the ANFIS controller is its robustness and

design simplicity. The ANFIS-PSS adjusts its parameters

adaptively on the basis of given input-output data to the

controller as training data.

B. ANFIS based PSS Design

An ANFIS-PSS is an artificial neural network based

approach which shows great potential of damping low

frequency electromechanical oscillations. Its design process

can be divided into two steps. Firstly, input-output data pair of

the system is generated using FLPSS which is called training

data. This training data determines the exact relation between

input and output of ANFIS-PSS. Then, a fuzzy inference

system (FIS) is created. FIS’s MF parameters tend to be

adjusted with the use of either only a backpropagation

algorithm, or in a permutation with a least squares method [18].

Gaussian MFs have been used here. After selecting number of

MFs, MF types, number of training epoch, and expectable

training error, the network is trained. Finally, if the FIS model

is valid, it is saved. Two inputs, rotor speed change (∆wr) and

rotor acceleration (∆wr´), and a Takagi-Sugeno type FIS model

are used in this paper. The inputs are Gaussian membership

type. Seven MFs are required to cover the full range of

respective inputs. In this way, 49 rules appear for the output

function which demonstrate a linear relation to the inputs.

The design procedure of ANFIS-PSS is as follows.

 Firstly, generate training data by using FLPSS.

 Then generate FIS system.

 Select number of MFs, types, epoch number and

maximum training error.

 Then start training and observe training error.

 Continue training until training error is within

expected level.

Then check model validation.

 Finally, save FIS structure.

R
t

0.4

0.3

0.2

0.1

0.0

-0.1

No PSS

AVR with PID

Root Locus

Freq. Resp.

FLPSS

ANFIS-PSS

2
4

10

Time(sec)

Figure 10. Response of rotor angle for a 5% change in mechanical input.

v
e

0.004

0.003

0.002

0.001

0.000

-0.001

-0.002

-0.003

-0.004
0

No PSS

AVR with PID

Root locus

Freq. Resp.

FLPSS

ANFIS PSS

10

Time(sec)

V. SIMULATION RESULTS

To analyze the performance of the designed PSS, a time-

domain simulation is carried out. The mechanical input to the

generator is increased (5%) at 1 sec. The response of rotor

angle and speed deviation of the synchronous generator for

different types of controller is shown in Fig. 10 and Fig. 11,

respectively. An AVR is suitable for regulation of generated

voltage but it introduces low frequency oscillations and

persists for a long period of time. Therefore, a PSS is needed.

The performance of root locus and frequency response PSS is

quite good. However,

they suffer from overshoot and

undershoot and take more time to settle down. It is to be

Figure 11. Response of rotor speed deviation for a 5% change in mechanical

input.

VI. CONCLUSION

This paper analyzes the performance of different types of

PSS design method. It is found that a generator without any

PSS has large angle and speed deviation. It is found that

conventional design method of PSS cannot guarantee

satisfactory performance under varying operating conditions.

Although fuzzy logic controller can enhance the system’s

performance, there are no definite criteria for selection of its

MFs. However, the designed adaptive PSS can overcome the

aforementioned difficulties. Simulation results show that the

801
ANFIS-PSS ensures the best performance in terms of setting

time and overshoot when operating condition changes.

Therefore, it can be effectively used for modern power system

applications.

APPENDIX

The system, input, output and feed forward matrices (A, B, C and D) are

given below [16].

A= [0 377 0 0 0 0 0

-.246 -.156 -.137 -.123 -.124 -.0546 0

0.109 0.262 -2.17 2.30 -.0171 -.0753 1.27

-4.58 0 30 -34.30 0 0 0

-.161 0 0 0 -8.44 6.33 0

-1.70 0 0 0 15.2 -21.5 0

-33.9 -23.1 6.86 -59.5 1.50 6.63 -114 ].

B= [0

0
 0

16.4].

C= [-.123 1.05 .23 .207 -.105 -.46 0

0 1 0 0 0 0 0

1.42 .9 .787 .708 .713 314 0].

D= [0

0].

Constants are given below.

K1=0.7636, K2=0.8644, K3=0.0331, K 4=1.4149, k5=-0.1463, k6=0.4146.

REFERENCES

[1] M. A. Mahmud, H. R. Pota, M. J. Hossain, and N. K. Roy “Robust

Nonlinear Controller Design

for Three-Phase Grid-Connected

Photovoltaic Systems Under Structured Uncertainties," IEEE Trans. on

Power Delivery, vol. 29, no. 3, pp. 1221-1230, 2014.

[2] P. Kundur, Power System Stability and Control. New York: McGraw-

Hill, 1994.

[3] H. Saadat, Power System Analysis. New York: McGraw-Hill, 1999.

[4] M. J. Hossain, M. A. Mahmud, N. K. Roy, and H. R. Pota,

“Enhancement of Transient Stability Limit and Voltage Regulation with

Dynamic Loads Using Robust Excitation Control,” International Journal

of Emerging Electric Power Systems, vol. 14, no. 6, pp. 561-570, 2013.

[5] D. K. Sambariya and R. Prasad, “Robust Power System Stabilizer

Design for Single Machine Infinite Bus System with Different

Membership Functions for

Controller,’’IEEE

Fuzzy

International Conference on Intelligent Systems and Control, pp. 13-19,

Tamil Nadu, India, 2013.

Logic

[6] R. Gupta, D. K. Sambariya , R. Gunjan, “Fuzzy Logic Based Robust

Power System Stabilizer for Multi-Machine Power System,” IEEE

International Conference on Industrial Technology, pp. 1037-1042,

Mumbai, India, 2006.

[7] G. Y. R. Vikhram and S. Latha,“Design of Power System Stabilizer

for Power System Damping Improvement using Optimization based

Linear Control Design” IEEE International Conference on Power

Electronics Drives and Energy Systems, vol. 8, pp. 2231-2307,

Bengaluru, India, 2012.

[8] R. You, H. J. Eghbali, and M. H. Nehrir, “An Online Adaptive Neuro-

Fuzzy Power System Stabilizer for Multimachine Systems,”IEEE Trans.

on Power System, vol. 18, no. 1, pp. 128-135, 2003.

[9] H. Ni, G. T. Heydt, and L. Mili, “Power System Stability Agents Using

Robust Wide Area Control,” IEEE Power Engineering Review, vol. 17,

no. 4, pp. 58, 2002.

[10] J. F. Ardanuy and P. J. Zufiria, “Adaptive power system stabilizer using

ANFIS and genetic algorithms,” IEEE Conference on Decision and

Control, pp. 8028-8033, Spain, 2005.

[11] P. Mitra, S. Maulik, S. .P. Chowdhury, and S. Chowdhury,“ANFIS

Based Automatic Voltage Regulator with Hybrid Learning Algorithm,”

IEEE 42nd International Universities Power Engineering Conference,

pp. 397-401, 2007.

[12] P. Mitra, S. Chowdhury, S. P. Chowdhury, S. K. Pal, Y. H. Song, and G.

A. Taylor,“Damping Local and Inter-Area Oscillations with Adaptive

Neuro-Fuzzy Power System Stabilizer,” Proceedings of the 41st IEEE

International Universities Power Engineering Conference, vol. 2. pp.

457-461, 2006.

[13] A. S. Venugopal, G. Radman, and M. Abdelrahman,“An Adaptive

Neuro Fuzzy Power System Stabilizer For Damping Inter-Area

Oscillations In Power Systems,”Proceedings of

the Thirty-Sixth

Southeastern Symposium on System Theory, pp. 41-44, 2004.

[14] V. K. Bhatt and S. Bhongade,“Design Of PID Controller In Automatic

Voltage Regulator (AVR) System Using PSO Technique,” International

Journal of Engineering Research and Applications, vol.3, no. 4, pp.

1480-1485, 2013.

[15] Ak. S.R. Swain, and A.K. Swain, and A. Mohapatra, “Design of Power

System Stabilizer” B. Tech. thesis, National Institute of Technology,

Rourkela, May, 2012.

[16] J. H. Chow, G. E. Boukarim, and A. Murdoch, “Power System

Stabilizers as Undergraduate Control Design Projects,” IEEE Trans. on

Power Systems, vol. 19, no. 1, pp. 144-151, 2004..

[17] N. Gupta, “Fuzzy Logic Based Power System Stabilizer,” Masters

thesis, Thapar University, Patiala, 2008.

[18] M. F. Othman, M. Mahfouf , and D. A. Linkens, “Designing Power

System Stabilizer for Multimachine Power System Using Neuro-Fuzzy

Algorithm,” jurnal teknologi, vol. 35, no. 1, pp. 55–64, 2001.

802