Fuzzy Based Power System Stabilizer

Abstract - A new type of Power System Stabilizer based on fuzzy set theory is proposed to improve the dynamic performance of a single machine power system. To have good damping characteristic over a wide range of operating conditions, &) of speed deviation ( ) and acceleration ( a machine are chosen as the input signals to the fuzzy stabilizer on the particular machine. These input signals are first characterized by a set of linguistic variables using fuzzy set notations. The fuzzy relation matrix, which gives the relationship between stabilizer inputs and stabilizer output, allows a set of fuzzy logic operations that are performed on stabilizer inputs to obtain the desired stabilizer output. In this the effect of variations of of defuzzification techniques on the performance of FLPSS (fuzzy logic power system stabilizer) has been investigated. Key words - PSS, Dynamic stability, fuzzy based PSS, Single machine infinite bus power system. I. INTRODUCTION Power System Stabilizer (PSS) have been extensively been used in modern power system for enhancing stability of the system. The PSS extends system stability limits by modulating generator excitation to provide damping to the oscillations of the synchronous machine rotors with respect to one another. The conventional Power System Stabilizer (CPSS), introduced nearly five decades ago, comprising a cascade connected lead-lag network with rotor speed deviation as input signal has made great contributions in enhancing stability of the system [1]. The conventional fixed structure PSS, designed using a laniaries model provides optimum performance for a nominal set of operating and system parameters. However, the performance becomes sub optimal following variations in system parameters and loading conditions from their nominal values. [2]-[4].In recent years, adaptive self tuning PSS genetic algorithm based PSS, artificial neural network based PSS and Fuzzy Logic based Power System Stabilizer (FLPSS) [3-5] have been proposed to provide optimum damping to the system oscillations under wide variations in system parameters and operating conditions. An adaptive self-tuning PSS suffers from a major drawback of requiring model identification in real-time, which is possible only with fast microcomputers. Although realization of a variable structure PSS is simple, it results in frequent change over of PSS structure and hence results in the problem of chattering. In ANN based PSS the ANN is trained by using a training set obtained from off line studies consideration the laniaries model of the system over the complete range of operating conditions. Thus, the knowledge of the system model is prerequisite for designing ANN based adaptive PSS. Fuzzy Logic controllers (FLC) are suitable for systems that are structurally difficult to model due to naturally existing non-linearity and other model complexities. Fuzzy Logic Power System Stabilizers unlike other power system stabilizers do not require a mathematical model of the system. In construct to a conventional PSS, which is designed in the frequency domain, a fuzzy logic PSS is designed in the time domain [4]. The fuzzy controllers determine the operating conditions from the measured values and select the appropriate actions from rules. Depending on the system state, the controllers operate in the range between no action and full action in an extremely nonlinear manner in order to damp out the oscillations. The fuzzy controller in it has no dynamic component, i.e. it can immediately perform the desired control action.

II.SYSTEM INVESTIGATED
A machine infinite bus system with synchronous generator provided with IEEE TypeST1 static excitation system is considered. Nominal parameters of the system are taken from ref [6]

III. DYNAMIC MODEL OF THE SYSTEM Following assumptions have been made in developing the dynamic model of the system. 1. The mechanical power input remains constant during the period of the transient. 2. Damping or a synchronous power is negligible. 3. The synchronous machine can be represented (electrically) by a constant voltage source behind a transient reactance. 4. The mechanical angle of the synchronous machine rotor coincides with the electrical phase angle of the voltage behind transient reactance. 5. If a local load is fed at the terminal voltage of the machine, it can be represented by a constant independence (or admittance) to neutral. The non-linear model of the system is obtained as follows
d (Tm Te ) = dt 2H (1) d = 2 f ( 0) dt (2) dv1 = ( Et V 1) / TR dt (3) d fd Efd = 2 f [ Rfd RfdLfd ] dt Ladu (4)

K1 KD 2H 2H 0 2 f 0 A = Where K 3K 4 0 T3 K5 0 TR , X = fd v1
1 2H = 0 , 0 0

K2 2H 0 1 T3 K6 TR

0 K 3 KA T3 1 TR 0

P= [ Tm ]

Fig1. Model for Constant Efd Fig. 2. Shows that transfer function of the system with PSS. The linear state space model is given in the form &= AX + p X (6)
KD 2H 2 f 0 A= 0 KSTAB.KD 2H T 1a 51 T2 K1 K2 2H 2H 0 0 K 3K 4 1 T3 T3 K5 K6 TR TR KSTAB.K 1 KSTAB.K 2 2H 2H T 1a52 T 1a53 T2 T2 0 0 K 3 KA T3 0 0 0 0 0 0 1 TR 1 TR 0 0 K 3 KA T3 0 0 1 T2

a linear model of the system is obtained by line arising the nonlinear model around a nominal operating point(transfer function model is given in the Fig. (1). The dynamic model of the liberalized system in the state space form is obtained from the transfer function model in the form: &= AX + p X (5)

T 1a55 +T2 T2

fd = v1 , v 2 ut
0 K 3 KA T3 0 0 0

1 2H 0 0 = KSTAB 2H T 1 KSTAB T 2 2 H

Tm p= vref

output signals are considered. These seven linguistic variables are PB (positive big), NS (negative small), NM (negative medium) and NB (negative big). A choice of 7 linguistic variables results in a set of 49 IF-THEN rules (Table.1) shows the decision table. This table is obtained from the expert knowledge of the operators. In order to obtain the minimum and the maximum values of the stabilizer inputs the dynamic performance of the system without PSS is obtained for different magnitudes of perturbations. After choosing the linguistic variables, it is required to determine the membership functions for these linguistic variables. Generally gauss Ian, triangular or trapezoidal membership functions are prevalent. Here triangular membership function is used to define the degree of membership (Fig. 3.). Degree of membership plays a very important part in designing a fuzzy controller.

Fig2. System Model with CPSS IV. FUZZY LOGIC PSS Selection of Input Signals to Fuzzy Logic PSS: For the present investigations generators &) are speed deviation ( ) and acceleration ( chosen as the input signals to the FLPSS. In practice, only shaft speed deviation ( ) is &) is readily available. The acceleration ( derived from the measured at two successive sampling instants ( KT ) [( K 1)T ] &(KT) = T Where T is the sampling period. Selection of Linguistic Variables: The number of linguistic variables determines the quality of the control, which can be achieved using Fuzzy Logic Controllers. As the numbers of linguistic variables increases, the quality of control improves at the cost of increased computational time and computer memory. A compromise is needed between the two. For the power system under study, seven linguistic variables for each of the inputs and the

Fig3.Symmetrical Triangular Membership function Now it is required to find the fuzzy region for the output for each fuzzy rule, for which Madman implication is used. Decision table1.

Fig4. System Model with FLPSS V COMPARISON OF PERFORMANCE OF THE SYSTEM WITH CPSS AND FLPSS Fig. 5 and 6 shows the transient response for and respectively following a 5 % step change in mechanical torque with conventional PSS and FLPSS obtained considering a linear model of the system. The scaling factors for and the stabilizing signals are respectively 5000, 350 and 500. Examination of the responses clearly shows that dynamic performance of the system with FLPSS is very close to that with conventional PSS and is in fact slightly better in terms of settling time. Further, in order to compare the dynamic performances quantitatively a quadratic performance index is evaluated J = [ (t )] = [ (kt )]2 t
2 0 k =0 .

Fig5.Speed deviation Vs Time Table 2. Shows the value of J for CPSS and FLPSS.it also shows the peak overshoot of the dynamic response for and . COG defuzzifier has been used and symmetrical triangular membership functions with equal base widths have been considered. Table2

Fig.6. Rotor angle deviation Vs Time

VI. EFFECT OF VARIATION IN COORDINATES OF MEMBERSHIP FUNCTION An attempt has been made towards gain tuning of the fuzzy logic controller by altering the coordinates of the membership functions. Center of Gravity / Area (COG/COA) defuzzifier is used in the analysis. Fig.7 shows a triangular membership functions, , , are the coordinates of the function. denotes the left coordinate of the base denotes the abscissa of the peak and denotes the right coordinates of the base. Let 1 , 2 and 3 be the X coordinates of the peaks of the membership functions PS, PM and PB respectively. For the membership functions PS let c1 = 1 For the membership functions PM let y = 2 1 For the membership functions PB let z = 3 - ( 1 + 2 ) =3-(y + c1 )

y z = .It can be clearly seen from Fig (9) that c1 y the coordinates of PS, PM and PB are (0, 1, 2), ( 1, 2, 3) and ( 2, 3,...) . 3 =3.0, the positive limit of the universe of discourse (universe of discourse is 3.0 to 3.0. The value of c1 is varied over a wide range in order to obtain its optimum value. Figs. (9) and Fig(10) show the transient response for and for a step increase in mechanical torque i.e. Tm = 0.05 pu for c1 = 1.2,1.0 and 0.4.Examination of the responses shows that the maximum peak overshoot is nearly the same for c1 = 1.2 and c1 = 1.0 and increases slightly for c1 = 0.4 whereas there is considerable reduction in setting time as c1 is reduced to 0.4 whereas there is considerable reduction in setting time for c1 is equal to 0.4.Examination shows that the performance index J decreases slightly while the settling time decreases to about 50% with change in c1 from 1.2 to 0.4.The change in peak deviation is quite insignificant. Fig. 9.Speed deviation Vs Time

Fig.7.

Fig.8.Trangular Membership Function At this stage, we assume that the separation between the peaks vary in geometric ratio i.e.

Fig.10. Rotor angle deviation Vs Time VII CONCLUSION A systematic approach for realizing Fuzzy Logic PSS has been presented. An attempt has been made to study the effect of variation of coordinates of the membership functions and scaling factors on performance of the FLPSS. Performance of the FLPSS has been compared with CPSS. Following are the significant results of the investigations carried out in this paper: 1. Performance of an optimum conventional PSS is similar to that of FLPSS. 2. Unsymmetrical triangular membership functions with c1 = 0.4 provides the best dynamic performance of the FLPSS. 3. The scaling factors decrease inversely with increase in magnitude of perturbation. 4. FLPSS using adaptive scaling factors provide good performance.