Australian Journal of Basic and Applied Sciences, 5(2): 258-263, 20 ISSN 99-878 A Study on Performance of Fuzzy And Fuzyy Model Reference Learning Pss In Presence of Interaction Between Lfc and avr Loops Mohammad Hossein Ferdowsi, 2 A.V.amyad, 3 halil Alizadeh Islamic Azad University, Gonabad Branch Gonabad, Iran Member of young researchers club. 2 Full professor, Department of Mathematics, Ferdowsi University of Mashhad. 3 Islamic Azad University, Gonabad Branch. Abstract: this paper reviews the previous studies on the effects of power system stabilizers (PSS) on damping generator s rotor oscillations through excitation control using auxiliary stabilizing signals. The authors of previous papers have neglected the interaction between two control loops: load frequency control (LFC) and automatic voltage control (AVR). Such assumption is, however, valid only in the conditions were amplitudes of disturbances are negligible or in special situations where the coupling between the two systems is not very strong. The present paper attempts to compare the performance of fuzzy logic PSS (FLPSS) with fuzzy model reference learning PSS (FMRLPSS) taking into account the interaction between LFC and AVR loops. The results show several advantages of fuzzy PSS over conventional PSS. ey words: Automatic voltage regulation; Load frequency control; Power system stabilizer; Fuzzy logic control INTRODUCTION The low-frequency oscillations are attributed to the oscillations of the mechanical mode of the system and can be approximately analyzed with a linear one-machine infinite bus model (Yau, N.A.N. yu 983). A complete system model for low-frequency oscillation studies must be included of mechanical and electrical loops. It has been recognized that these oscillations can be controlled by adusting exciter and speed-governor control parameters. Furthermore, it has been shown that the load-voltage characteristic of the power system has a significant effect on its dynamic responses, and suggestions have been made for the proper representation of these characteristics in simulation studies (S.C. Tripathy, N.D. Rao and L. Roy 98;. Yamashita and H. Miyagi 99). The two main control loops of a generation are Load Frequency Controller (LFC) and Automatic Voltage Regulator (AVR) as seen in Fig.. The turbine fed by controllable rate of steam and the Automatic Generation Control method deals with frequency through the LFC loop and with voltage through with the AVR loop, where the main purposes of these controllers are to maintain levels of voltage and frequency at the acceptable values. However, these studies are based on the assumption that there is no interaction between the power/frequency and the reactivepower/voltage control loops. This assumption is permissible only when the speed of the excitation system is much faster than that of the LFC system; in the practical systems, during dynamic perturbations, exists some interaction between these two control channels (E. Rakhshani,. Rouzbehi and S. Sadeh 2009). Fig. : Automatic generation control with LFC and AVR loops. Due to the weak coupling relationship between the AVR and LFC, the voltage and frequency are regulated separately. The study of coupling effects of the AVR and LFC loops can be found in (N. Jaleeli et al., 992; Ibraheem, P. umar and D.P. othari 2005). But in all of these researches there are sketchy attentions to these mutual effects. Furthermore there is no any attention to the turbine output power in the steady state. Also, by neglecting the effect of voltage deviation on load demand, an important interaction in LFC systems is ignored. In order to prevent this problem and improve the accuracy of responses, Rakhshani and et al., in (2009). proposed a Corresponding Author: Mohammad Hossein Ferdowsi, Islamic Azad University, Gonabad Branch Gonabad, Iran Member of young researchers club. E-mail: Hosseinferdowsi@gmail.com 258
Aust. J. Basic & Appl. Sci., 5(2): 258-263, 20 combined classical model for low frequency oscillation studies. In this paper this model is used to compare FLPSS and FMRLPSS performance in different conditions (see Fig. 2). Fig. 2: Proposed combined model for compare the performance of CPSS and FLPSS. Problem Formulation: System Modeling: In single machine infinite bus system, the synchronous machine (generator) is connected to an infinite bus through a transformer and two parallel transmission lines. In generator bus, a local load is also supplied as seen Fig. 3. Fig. 3: Schematic of the single machine power system connected to an infinite bus. Fuzzy Logic Power System Stabilizer (FLPSS: Power systems usually operate under highly uncertain stress condition. Moreover, load changes cause the variation of the generator dynamic characteristics so that the different operating conditions are obtained. Therefore, power system controllers should be designed to maintain the robust stability of the system. On the other hand, a CPSS is designed for a linear model representing the generator at a certain operating point and it often does not provide satisfactory results over a wide range of operating conditions. To overcome these drawbacks, fuzzy logic controller (FLC) is an effective tool, which has nonlinear structure. In fuzzy controller design, there is no need to perfect model of the system, which is significant advantage. In what follows, we will describe how the FLPSS has been synthesized. The design process of fuzzy logic controller maybe split in to five steps: ) the selection of control variables, 2) the membership function definition or the fuzzification, 3)the rule creation or the knowledge base, 4) the fuzzy inter face engine, and 5)the defuzzification strategy or the defuzzifier. These steps are shown in Fig. 4. Fig. 4: The basic structure of the fuzzy controller. Also, in Fig. 5 it is shown how to use fuzzy controller in a PSS structure and its illustrations can be explained as the following steps (R. Hooshmand and M. Ataei 2009). 259
Aust. J. Basic & Appl. Sci., 5(2): 258-263, 20 Fig. 5: Schematic structure of FLPSS. Step (): In this method, two variable and are used as input signal in PSS. The coefficient in and in2 in input stage, keep the input signals within value scale to required value in decision limit. The output signal (U PSS ) is inected to the summary point of AVR as the supplementary signal. Step (2): Each of FLPSS input and output fuzzy variable Y=(, U PSS ) membership function have been chosen identical because of the normalization achieved on the physical variables. The normalization is important because is allows the controller to associate equitable weight to each of thr rules and therefore, to calculate correctly the stabilizing signal. Each of the input and output fuzzy variable, y i is assigned seven linguistic fuzzy subsets varying from Negative Big (NB) to Positive Big (PB). Each subset is associated with a triangular membership function to form a set of seven normalized and symmetrical triangular membership function for fuzzy variables. (see Fig. 6). Fig. 6: Fuzzy variable ya seven membership functions. Step (3): The y max and y min represent maximum and minimum variation of the input and output signals. These values are selected based on simulations data. The range of each fuzzy variable is normalized between -4 to 4 by introducing a scaling factor to represent the actual signal. Step (4): The interface mechanism of the FLC is represented by a 7 7 decision table. The set of decision rules relating all possible combinations of input to outputs is based on previous experience in the field. This set is made up of 49 rules expressed using the same linguistic variables as those of the inputs and is stored in the form of a decision table shown in Table I. Table : flpss decision table. NB NM NS Z PS PM PB NB NB NB NB NB NM NS Z NM NB NB NM NM NS Z PS NS NB NM NM NS Z PS PM Z NM NM NS Z PS PM PM PS NM NS Z PS PM PM PB PM NS Z PS PM PM PB PB PB Z PS PM PB PB PB PB Step (5): Let, 2,..., 3 represent the centroids of M membership functions that are assigned to U PSS and w i represents the firing strength of the ith rule. Thus, for M rules, the output of the controller is: 260
Aust. J. Basic & Appl. Sci., 5(2): 258-263, 20 U PSS Where M w. J M w. [,,, and 2 M ] i M w i w (3) Fuzzy Model Reference Learning PSS: Figure 7 shows the functional block diagram of the FMRLPSS. It is made up of four main parts; the plant, the fuzzy controller to be tuned, the reference model, and the learning mechanism (an adaptation mechanism) (Layne.R.,.M. Passino, 996). The FMRLPSS uses discrete time signals (r(, and y( with T as the sampling period. It also uses the learning mechanism to observe numerical data from a fuzzy control system. With this numerical data, it characterizes the fuzzy control system s current performance and automatically synthesizes or adusts the fuzzy controller so that some given performance obectives are met. These performance obectives, which is the closed loop specifications are characterized through the reference model of Fig. 7. Fig. 7: Fuzzy Model Reference Learning PSS. Here, the fuzzy control system loop operates to make y( track r( by manipulating u(, while the adaptation control loop seeks to make the output of the plant y( track the output of the reference model ( by manipulating the fuzzy controller parameters. y m C. The Fuzzy Controller: The synchronous generator which represents the plant has an input u( from the fuzzy controller and terminal voltage output y(. The input to the fuzzy controller is the error e( r( y( and change in e( e( kt T ) error c( T Where r( is a reference input. A total of 2 fuzzy rules were employed as indicated below in table with triangular membership functions. Table 2: fmlrpss decision table. NV NL NB NM NS Z PS PM PB PL PV NV NV NV NV NV NV NV NL NB NM NS Z NL NV NV NV NV NV NL NB NM NS Z PS NB NV NV NV NV NL NB NM NS Z PS PM NM NV NV NV NL NB NM NS Z PS PM PB NS NV NV NL NB NM NS Z PS PM PB PL Z NV NL NB NM NS Z PS PM PB PL PV PS NL NB NM NS Z PS PM PB PL PV PV PM NB NM NS Z PS PM PB PL PV PV PV PB NM NS Z PS PM PB PL PV PV PV PV PL NS Z PS PM PB PL PV PV PV PV PV PV Z PS PM PB PL PV PV PV PV PV PV 26
Aust. J. Basic & Appl. Sci., 5(2): 258-263, 20 In the table above, NV, NL, NB, NM, NS, ZR, PS, PM, PB, Pl, PV stands for negative very large, negative large, negative big, negative medium, negative small, zero, ositive small, positive medium, positive big, positive large, and positive very large. Fig. 8: Membership functions for input universe of discourse. Fig. 9: Membership functions for output u. C2. The Reference Model: A reference model G ( s) is chosen because this model decays to zero in short time. If T = 0. sec, we s can use bilinear transformation to find the discrete equivalent continuous time transfer function G(s) by replacing 2 z s with A Fuzzy Model Reference Learning PSS for Synchronous Generator Terminal Voltage Control T z ( z ) ( z) H ( z) 2 R( z) 9 z 2 y m where y m (z) and R(z) are the transforms of y m (kt ) and r(kt ) respectively. So the discrete time implementation is 9 ym( kt T ) ym ( r( kt T ) r( kt ) 2 2 2 C3. The Learning Mechanism: The learning mechanism tunes the rule-base of the direct fuzzy controller so that the closed loop system behaves like the reference model. These rule-base modifications are made by observing data from the controlled process, the reference model, and the fuzzy controller. The learning mechanism consists of two parts: a fuzzy inverse model and a knowledge base modifier. The fuzzy inverse model (having the same rule base with the fuzzy controller) performs the function of mapping y e ( (representing the deviation from the desired behavior) to changes in the process inputs p( that are necessary to force y e ( to zero. The knowledge-base modifier performs the function of modifying the fuzzy controller s rule-base to affect the needed changes in the process inputs. simulation results: In this section, in order to compare the performance of the FLPSS and FMRLPSS, some simulations are performed and its time domain results are provided. Simulations performed in three different operating conditions as follow: a) Normal load condition: The condition in which the system is operated in initial values. The values are selected as P e0 =.0 p.u, Q e0 =0.05p.u, V t0 =.05pu. b) Heavy load conditions: The condition in which the real power (P e ) is increased from.0 to.3 p.u. 3) In the case of fault occurrence in transmission line: The condition in which the line 2 in Fig. is isolated with normal condition. 262
0-0.5 - -.5-2 Time (s) Time (s) FMRL PSS FL PSS 0 2 3 4 5 6 7 8 9 0-2.5 Aust. J. Basic & Appl. Sci., 5(2): 258-263, 20 Table 3: the coefficients k to k 6 for the heffron-phillips model in different operational conditions. Operation Conditions 2 3 4 5 6 Nominal Load 0.544.2067 0.6584 0.698-0.095 0.859 Heavy Load 0.4563.4477 0.6584 0.8706-0.67 0.7747 Fault in the Line 0.4007.404 0.7095 0.6834-0.20 0.8348 In order to compare the performance of FLPSS and FMRLPSS, the load change in real power is set at 0% and the behavior of frequency deviation in different operational conditions are shown in Figs. 7(a)-7(c). 2 x 0-4 0 Frequency deviation (HZ) - -2-3 -4 FMRL PSS FL PSS -5 0 2 3 4 5 6 7 8 9 0 2 x 0-4 (a) 0-2 Frequency deviation (HZ) -4-6 -8-0 FMRL PSS FL PSS 0.5 x 0-3 -2 0 2 3 4 5 6 7 8 9 0 (b) (c) Fig. 0: Compare performance of FLPSS and FMRLPSS: (a) Normal Load, (b) Heavy Load, (c) Fault in the line. Conclusion: Several studies have shown that LFC loops are not completely decoupled from AVR loops. In fact, the interaction between the two loops may only be neglected when disturbances are small in magnitude or coupling coefficients are small. In this paper, performance of fuzzy PSS was compared to that of fuzzy model reference learning PSS in presence of interaction between LFC and AVR loops. The simulation results suggest that fuzzy model reference learning PSS outperforms fuzzy logic PSS in different working conditions. REFERENCES Hooshmand, R., and M. Ataei, 2009. An Auto-Tuning Fuzzy Logic PSS Design under Multi-operating Conditions Using Real-Coded Genetic Algorithm" Journal of Electrical Systems, vol. 5, no.. Ibraheem, P., umar and D.P. othari, 2005. " Recent Philosophies of Automatic Generation Control Strategies in Power Systems,'' IEEE Trans. on Power Systems, 20(): 346-357. Jaleeli, N., L.S. VanSlyck, D.N. Ewart, L.H. Fink and A.G. Hoffmann, 992. "Understanding automatic generation control,'' IEEE Trans. on Power Systems, 7(3): 06-2. Layne J.R.,.M. Passino, 996. Fuzzy Model Reference Learning Control, Journal of Intelligent andf uzzy Systems, 4(): 33-47. Rakhshani, E.,. Rouzbehi and S. Sadeh, 2009. A New Combined Model for Simulation of Mutual Effects between LFC and AVR Loops.IEEE Conf. Power and Energy Engineering. pp: -5. Tripathy, S.C., N.D. Rao and L. Roy, 98. ''Optimization of exciter and speed governor control parameters in stabilizing intersystem oscillations with voltage dependent load characteristics,'' Electric Power and Energy Systems, 3: 27-33. Yamashita,. and H. Miyagi, 99. ''Multivariable self-tuning regulator for load frequency control system with interaction of voltage on load demand,'' in Proc. IEE Control Theory and Applications Conf., 38: 77-83. Yau, N.A.N., yu, 983. Electric Power System Dynamics, London, Academic Press. 263