Journal of Artificial Intelligence ISSN 1994-5450 science alertJournal of Artificial Intelligence 5 (1): 26-36, 2012 ISSN 1994-5450 / DOI: 10.3923/jai,2012.26.36 © 2012 Asian Network for Scientific Information Fuzzy Logic Controller Optimization Based on GA for Harmonic Mitigation ‘Rachid Dehini, *Brahim Ferdi and 'Benaissa Bekkouche ‘Department of the Sciences and Technology, University (Center) of Relizane, Algeria "Department of the Sciences and Technology, University of Oran, Algeria Corresponding Author: Rachid Dehini, Department of the Sciences and Technology, University (Center) of Relizane, Algeria ABSTRACT In this study, a pulse generation method to trigger the MOSFETs of a shunt active power filter (SAPF) inverter is examined. The proposed method is aimed to compensate current. harmonic, as well as reactive power compensation, in power systems with a large concentration of non-linear loads. Design of optimal fuzzy logic controller by genetic algorithm method is presented. Some parameters defining the fuzzy controller (membership functions, fuzzy rules) are optimized simultaneously. In order to investigate the performance of this control method, all studies have been carried out using simulation with the MATLAB Simulink power system toolbox, The results of simulation study of new SAPF control technique presented in this paper are found to be quite satisfactory by assuring good filtering characteristics and high system stability. Key words: Genetic algorithm, optimal fuzzy controller, membership functions, fuzzy rules, shunt active filter, total harmonic distortion Nowadays, with the widespread increase of power electronic loads in industry, significant non- linear loads appear in a considerable amount. in harmonic injection and low power factor of power systems. They tend to introduce harmonies in voltage and current waveforms at the point of common coupling. The existence of current and voltage harmonies in power systems increases lines losses, decreases the power factor, and can cause timing errors in sensitive electronic equipments (Lopez et al., 2003; El-Mofty and Youssef, 2001) Rapid improvement in power semiconductor device makes high-speed, high-power switching usable for the harmonic compensation modern power electronic technology (Ahmed, 2004). Active Power Filters (APF) have been considered as an effective solution for this issue and have been widely used, ‘The SAPF working is to sense the load currents and extract. the harmonie component of the load current to produce a reference current (Liu et al., 2008). A block diagram of the system is illustrated in Pig. 1 To control SAPF, many techniques are used, among them the classieal PI controller and the relatively recent one the fuzay controller. It is difficult to get satisfied control characteristies by using a normal linear proportional plus integral (PD) controller due to the nonlinearity of the system (SAPF + fast nonlinear load variations). Also, fuzzy controller design is not obvious process due to the fact that there is not a known systematic design techniques presently exist. Although the 26ol. Artif. Intel., 5 (1): 26-36, 2012 Fig. 1: Schematic diagram of shunt SAPF funetions of fuzay systems have the advantage of being relatively easy to understand, the systems become complex because more design parameters are required such as number of input variables, number of fuzzy sets, rules, membership functions and the input and output. gains. It is possible to build a fuzzy controller with parameters determined by intuition or experience, which provides better performance than a conventional PI controller, but it may not be exactly an optimal one. Since the optimal system is generally the goal of controller researchers, this work is an attempt. to undertake a systematic study of the problem to provide some design guidelines for the optimal design of a fuzzy controller based on Genetic Algorithm (GA) (Berbaoui et al., 2010; Wang and Kwok, 1992). The proposed controller is applied to Shunt Active Power Filter (SAPF) current control with fast nonlinear load variations, METHODS Control by FLC: As illustrated in Fig. 2, the input of the FLC is e(t) and Ae(t). y (&) is the perfect output of the FLC (Dixon et al, 1999), I, (t) is the actual output of the SAPF: e(= At) - 1, @) @ The output of the FLC y (t) is used over the reference eurrents (ira, irb, ire) and sensed load currents (ila, ilb, ile) to generate switching signals for the MOSFETs of inverter. In response to switching signals generated by controller, the SAPF shapes the supply currents to sinusoidal and compensates the power factor, current harmonics and the unbalance of the nonlinear load, Mathematical formulation of the problem: The amount of distortion in the eurrent waveform is quantified by means of an index called the total harmonic distortion (THD). The THD in current is defined as: speo=to0e f° [ ® mal 8ol. Artif. Intel., 5 (1): 26-36, 2012 EAE ve ZA) E> uy logis controller “J i Ly & Fa [E> eo SD ¢ Tey og ene Li ened caret oA rs a Fig. 2: The optimum fuzzy controller where, I,, represents the RMS value of rank n harmonic component. I,, represents the RMS value of fundamental component. Interpreting the THD of the current supply of each phase associated with such FLC design used for pulse generation. The objective function is presented i f(@)=THD, 8 ‘The problem is to minimize multiple objective functions under the following constraints: a) where, dx, distance in Membership functions, as shown in Fig. 3. For the application of genetic algorithm, we use the minimization method under constraints, which is the method of penalty @20150...m @ We transform the problem into a penalty function which is presented as follows: feqan)= SoH / where, 1, is the coefficient of penalty. Optimization of fuzzy controller parameters using GA: The operation of a FLC depends on a large number of parameters (functions membership, fuzzy rules, rules inference, defuzzification) 28ol. Artif. Intel., 5 (1): 26-36, 2012 Fig. 3: Membership functions using the apexes differences Fig. 4: Hierarchical structure of the GA approach for optimized FL which must be settled at the design. Since these parameters influence each other, it is unlikely that a synthesis method dealing separately with each FLC subsystem can provide optimal results (Wang and Kwok, 1992). Several design methods for specifying the different parameters for the functioning of a fuzzy system have been deseribed in literature with the aim of ensuring the best usefulness of parameters for a given controller structure. In this section, we present the automatic method for the designation of fuzzy systems that enable to determine the optimal fuzzy system; this method is shown in Fig. 4 Membership functions: In an FLC, the reasoning rules ave applied to the linguistic terms. These terms, which can specify a linguistic variable are defined via membership functions. There are various methods for the implementation of membership functions. In our case we use adjacent triangular membership functions and show how to present it on a chromosome using binary coding. Figure 3 represents a fuzzy partition defined by seven triangular membership functions. This type of linguistic variable has been proposed by Yubazaki et al, (1995). A partition with n fuzzy linguistic terms is completely defined by n points (x1 = a, x2,....xn = b) using the triangular membership functions as follows:ol. Artif. Intel., 5 (1): 26-36, 2012 SYN fy X Rha The rules represented by a binary string, in which each element gi , takes its values in the set {1...N,3, being coded on u bits where: ve a =F(logy() +D) Bits aay where, f(x) is the integer part of x As shown in Fig.Gaandb, the total number of parameters to eneode of the fuzzy system is 21 (7 differences of membership functions for each input and output) +49 conclusion of a ruler), For the representation of the simple AG, the differences between the apexes are coded to 08 bits and the i, are encoded to 08 bits RESULTS AND DISCUSSION ‘The performance of the proposed synthesis FLC method was examined through simulations ‘The system model was implanted in MATLAB Simulink environment, The SAPF was designed to compensate harmonies caused by nonlinear loads. The system model parameters are shown in Table 1 A three-phase diode rectifier with an RL load was used as @ harmonic producing load, The initial load (resistance was 10/3 Q and the inductance 60 mH.) In the first place, the load power were raised with the following perosntage 25, 50, 75% at 0.2, 0.4, 0.8, then the total harmonic distortion has been taken up to 2.5 kHz for each case, A phase-a load current waveform is presented in Fig. 7 For the defuzzification operation, the center of gravity (also known as centroid) is used; this method allows us to express analytically the output of fuzzy system, to simplify implementation and also to reduce the computing time. The output expressing in this method given by equation: (0) dx a5 fox ax The seven values of difference membership function of each input and the output are encoded and grouped in a single chromosome, as shown in Table 2 ‘The GA which provides optimization of fuzzy controller is used only during the FLC synthesis phase, It is no longer involved in the system during normal operation. The parameters chosen for GA are: Initial population = 25; Selection function: Rolette; Crossover function: Two point; Crossing probability = 0.6; Mutation function: uniform; Mutation probability = 0.001, Maximum generation 0 number Baol. Artif. Intel., 5 (1): 26-36, 2012 ‘Table 1: System parameters Aetive filter Parameters Supply phase veltage U B0V Supply frequency fe He DC link eapacitor Ct 0768474 mE ‘Table 2 Membership functions differences of each FLC variable Input Coding Constraint ° 0 o.s18 02402 oeee 0.2619 00201 0.209 els be 0 0.2107 024 0.4250 0.2891 04000 0.2566 S422 = output neat Uref PWM 0 0.2519 ose8 0.2407 0.163 0.1958 0.3880 Sa was Table 3: Fuzzy rules ofthe optimum FLC se NB XM NS a 8 PM Pa NB NB NB NB NB NM NS Ze NM NB NB NB NM Ns a Ps xs xB xB xM Ns ze Ps Pw 2 xB xa xs ze "8 Pu Pa Ps nM Ns 2B Ps PM iB PB PM NM me Ps Pu "B » Pa Pe xs 8 PM in mB me Pa 200 E 0 3 0 = aod 2 00) 0 01 02 03 04 05 06 07 08 09 10 Time(s) Fig. 7; Phase‘a’ load current waveform ‘The differences in membership funetions for each variable of FLC resulting data are illustrated in Fig. 8 and Table 8. The result of simulation makes clear that the increase of the load current value yield to exceed of the harmonic in supply current or the conventional controller can not react enough against the variation of load, as shown in Fig. 11 BBol. Artif. Intel., 5 (1): 26-36, 2012 BS ip ogres of y a © Inpa 2 (ef) Degree of membership 06 04 024 of “de 0s 4 42 0 02 0” 06 G8 © Ourpet 1 Fig. 8: Membership function of (a) e(k) input (b) Ae(s) input ()the output, CAGLIARI ITER I Phase (injected cree ‘waveform 100 Time ae) “haa Hitt Pea) spt current veavefom, oa Time (ee) Fig. 9(a): Phase'a' injected current waveform,(b) phase ‘a’ supply current waveform with the(CP1) According to Fig. 5, it is observed that the genetic algorithm which designing the Fuzzy controller decrease the THD when the number of iterations increases. The filtering result ean be Bdol. Artif. Intel., 5 (1): 26-36, 2012 Pinee (9 injected current veaveform woo LE i ‘waveform Phase) supply curent 200) a a ‘Time eee) Fig. 10(a, b): (a) phase'a’ injected current, waveform, (b) phase ‘a’ supply current waveform with the optimum FLC 19 ! z : Fe eon Time(seo Fig. 11: Supply current THD, comparison between optimum FLC and CPI seen in Fig. 9-11. The deformations have now been reduced and the harmonic has been weakened with the optimum FLC where the THD calculated up to 2.5 kHz remains less than the case of conventional PI controller throughout the growth of the load. CONCLUSION With the variation of non-linear loads, the harmonics and reactive power are deteriorating the performance of the utility network. In this paper, a synthesis method of fuzzy controller parameters is presented to substitute the supplied analysis by a human expert, The controller designed according to fuzzy logic rules and the optimal adjustment of the membership functions are realized by using GA optimization FLC. A.control strategy based on FLC/GA for high performance three-phase SAPF has been proposed and confronted with CPI controller. The efficiency of the proposed FLC was verified through BBol. Artif. Intel., 5 (1): 26-36, 2012 simulation studies with MATLAB. FLC/GA has an important effect on the convergence of the problem to best solutions at load changing and also at the steady state and can act positively on performance of SAPF. FLCIGA is also found extremely satisfactory to mitigate harmonies and reactive power components from utility eurrent, REFERENCES Ahmed, M., 2004. Sliding mode control for switched mode power supplies. D.Se, Thesis, Department of Electrical Engineering, Lappeenranta University of Technology, Lappeenranta. Berbaoui, B., B. Chellali, R, Dehini and B, Fardi, 2010. Optimization of shunt active power filter system fuzzy logic controller based on ant colony algorithm. J. Theor. Applied Inform. Technol., 14: 117-126. Dixon, J.W.,J.M. Contardo and L.A. Moran, 1999. A fuzzy-controlled an Active front-end rectifier with current harmonic filtering characteristics and minimum sensing variables. IEEE Trans Power Electron., 14: 724-729. El-Mofty, A. and K. Youssef, 2001. Industrial power quality problems. Proceedings of the 16th International Conference and Exhibition on Electricity Distribution, Part 1: Contributions, June 18-21, 2001, CIRED, Amsterdam, Netherlands, pp: 5. Liu, H., G. Liu and Y. Shen, 2006. A novel harmonics detection method based on wavel et al gorithm for aetive power filter. Proceedings of the 6th World Congress on Intelligent Control and Automation, June 21-23, 2006, Dalian, China, pp: 4479-4483. Lopez, M., L. Moran, J. Espinoza and J. Dixon, 2003, Performance analysis of a hybrid asymmetric multilevel inverter for high voltage active power filter applications. Proceedings of the IEEE Industrial Electronics Conference, November 26, 2003, Roanoke, Virginia, USA, pp: 1050-1055, Wang, P. and D.P. Kwok, 1992, Optimal fuzzy PID control based on genetic algorithms Proceedings of the International Conference on Industrial Electronics, Control, Instrumentation and Automation, November 9-13, 1992, IEEE Press, San Diego, pp: 977-981 Yubazaki, N., M. Otani, T. Ashida and K. Hirota, 1995, Dynamie fuzzy control method and its applications to positioning of induction motor, Proceedings of 1995 IEEE International Conference on Fuzzy Systems, International Joint Conference of the 4th IEEE International Conference on Fuzzy Systems and The 2nd International Fuzzy Engineering Symposium, March 20-24, 1995, Yokohama, Japan, pp: 1095-1102,