Adaptive Neuro-Fuzzy Inference System Based Active Power Filter for Power Quality Improvement

Transcription

1 AUSTRALIAN JOURNAL OF BASIC AND APPLIED SCIENCES ISSN: EISSN: Journal home page: Adaptive Neuro-Fuzzy Inference System Based Active Power Filter for Power Quality Improvement 1 J.Jeraldin, 2 M.Arokiamary, 3 M.Vennila 1 P.G.Scholar, 2 Assistant Professor, 3 Assistant Professor Address For Correspondence: J.Jeraldin, P.G.Scholar jeraldinjoseph@gmail.com; A R T I C L E I N F O Article history: Received 12 January 2016 Accepted 22 February 2016 Available online 1 March 2016 Keywords: Shunt active power filter, fuzzy logic, ADALINE, membership function, harmonic compensation. A B S T R A C T The main objective of this paper is to develop a new method for reactive power compensation and harmonic compensation using ADALINE and a new control algorithm for active power filter with different membership function to improve power quality. Shunt active power filter is a complete current source which provides an adjustable and effective solution for current harmonic elimination in electrical power systems. Whenever the fuzzy membership functions are triangular and trapezoidal, the supply voltages are sinusoidal and balanced which is converging to compensation and when the Gaussian membership function occurs supply voltages are non sinusoidal and unbalanced. This paper simulates hybrid active power filter where active part is used to filter out the higher order harmonics and the passive part is used to filter out lower order harmonics. The shunt active power filter performance was evaluated with MATLAB/SIMULINK under balanced voltage conditions. The simulation results are obtained to mitigate the harmonics of the shunt active power filter with fuzzy logic control. INTRODUCTION Generally, now a days, houses with non linear loads occupying 80% of the total loads, where industrial devices are non linear or become non linear when combined with additional control circuits. Harmonic produces voltage and current waveforms which are highly offered by non linear characteristics in electrical distribution system. In Power distribution system non linear loads appear to be prime source for harmonic distortion. According to the concept of point of common coupling, non linear loads are injected back to power distribution systems which produce harmonic currents. Active power filter(apf) are advanced solutions and one of the most appropriate solutions for compensating for harmonics in power distribution networks.apf allows compensation of voltage harmonics (series APF) and compensation of current harmonics (shunt APF) with reactive power compensation required by non linear load.ann (artificial neural network) has been used as efficient methods to control non linear and complex systems, to estimate harmonic components based on back propagation learning rule. In presence of noise, it leads to inaccurate results because of data training. Under non ideal conditions this control scheme is complex and difficult. Neuro fuzzy system combines the learning capabilities of neural networks and the control capabilities of fuzzy logic system. The fuzzy controller is able to improve the performance by adjusting the current error via fuzzy rules. To improve the power quality by mitigating the harmonics of shunt active power filter, fuzzy logic control with different membership function is developed. ADALINE (adaptive linear neuron) estimates the current representative of the fundamental frequency per phase on-line learns. In order to compensate the harmonics through the point of common coupling it is able to re-inject the adequate reference currents into the power Open Access Journal Published BY AENSI Publication 2016 AENSI Publisher All rights reserved This work is licensed under the Creative Commons Attribution International License (CC BY). To Cite This Article: J.Jeraldin,.Arokiamary,Vennila., Adaptive Neuro-Fuzzy Inference System Based Active Power Filter for Power Quality Improvement. Aust. J. Basic & Appl. Sci., 10(5): 59-68, 2016

2 60 J.Jeraldin et al., 2016 distribution networks. Fundamental currents are synchronized with the direct components of the voltage issued from PLL (phase locked loop). In order to maintain the power factor compensating of the phase shift are done between the fundamental currents and the direct components of the voltage in order to maintain the power factor. The fuzzy controllers does not require an accurate mathematical model which can work with imprecise input and can handle non linearity and robust than conventional controllers. ANFIS (adaptive neuro fuzzy inference system) are class of adapting networks that are functionally equivalent to fuzzy inference systems. It uses hybrid learning algorithm. Using the given input or output data set, the toolbox function ANFIS constructs a fuzzy inference system (FIS) whose membership function parameters are turned using either a back propagation algorithm alone or in combination with least square type of method. This adjustment allows fuzzy systems to learn from data they are modeling. The fuzzy logic control (FLC) with trapezoidal membership function is unable to suppress the harmonics efficiently. The fuzzy logic control with triangular membership function (M.F) gives better performance than FLC with trapezoidal membership function, but, the FLC with Gaussian gives an outstanding performance than FLC with trapezoidal and triangular under any voltage conditions. Through MATALB simulation fuzzy logic control are verified effectively. II. SYSTEM CONFIGURATION A. Shunt Active Power Filter To compensate reactive power for linear load /non linear loads, shunt active power filter is used to eliminate the harmonic currents and to determine the desired compensation currents APF (active power filter) senses the source voltages and load currents. Active filters are implemented using a combination of passive and active components and require and outside power source. Fig. 1: Shunt Active Power Filter The goals of the shunt active power filter control are unity source power factor at positive sequence fundamental frequency, minimum average real power consumed or supplied by the active power filter and harmonic and neutral current compensation. To eliminate the harmonic and reactive components of load currents resulting in sinusoidal and unity power factor source currents active filter is observed. AF (active filter) enhances the system efficiency because the source need not process the harmonic and reactive power demanded by the load. Results obtained by simulations with MATLAB/SIMULINK is more effective than the reviewed approaches on compensating reactive power and harmonics of the load, even if the source voltages are severely distorted and imbalanced. B. Fuzzy Logic Technique: A fuzzy inference system for fuzzy system consists of a formulation of the mapping form a given input set to an output set using fuzzy logic. A fuzzy inference process consists of following steps; Step1: Fuzzifier the input variables Step2: Applying fuzzy operator (AND, OR, NOT) in the IF part of the rule. Step3: Implication from the antecedent to the consequent (THEN) part of the rule. Step4: Aggregation of the consequents across the rules. Step5: Defuzzifier To represent knowledge; the most common way is to represent human knowledge in the form of natural language expressions of the type. IF premise (antecedent) THEN conclusion (consequent). IF-THEN rule based form is commonly referred in this form of expression. This form of knowledge representation is quite appropriate in the context of linguistic because it express human empirical and heuristic knowledge in our own language of communication. Aggregation of rule is defined as the process of obtaining the overall consequent from the individual consequents contributed by each rule in the rule based.

3 61 J.Jeraldin et al., 2016 The Basic Structure of Fuzzy Systems is fuzzifier, inference engine and defuzzifier. The Fuzzifier is Converting the crisp input to a linguistic variable using the membership functions stored in the fuzzy knowledge base, the Inference engine is Using IF-THEN type fuzzy rules converts the fuzzy input to fuzzy output and Defuzzifier is Converting fuzzy output of the inference engine to crisp using membership functions analogous to the ones used by the fuzzifier. Fig. 2: Basic Structure of Fuzzy Systems III. Proposed Control Technique: A. ANFIS Technique: Fig. 3: ANFIS block The adaptive neuro fuzzy inference system technique is a controlling unit depends upon the main subsystem where the system consists of 8 inputs and 6 outputs. The input 1, 3, 5 are the injecting current and 2, 4, 6 are the load current whereas, the In7 represents the 3phase supply of 415V and In8 represents the dc link voltage. The outputs from the subsystem are S1 to S6 representing pulse generation from MOSFET. Inputs are taken from IN7 and In8 for error and a memory block is kept for change in error. The input 1, 3, 5 are the injecting current and 2, 4, 6 are the load current whereas, the In7 represents the 3phase supply of 415V and In8 represents the dc link voltage. The outputs from the subsystem are S1 to S6 representing pulse generation from MOSFET. Inputs are taken from IN7 and In8 for error and a memory block is kept for change in error.

4 62 J.Jeraldin et al., 2016 Fig. 4: Subsystem 2 of the ANFIS Technique Fig. 5: Subsystem 4 of the ANFIS Technique Fig. 6: Subsystem 1 of the ANFIS Technique B. Membership Function: To identify the membership function parameters of single output ANFIS uses hybrid learning algorithm. Membership functions are the building blocks of fuzzy set theory. The FIS variables in the membership function editor are error, change in error and output. The maximum supply voltage is 415V divided into 7 membership function as NB(negative big), NM(negative medium), NS(negative small), zero, PS(positive small), PM(positive medium), PB(positive big). A membership function can have different shapes. The commonly used membership function is the triangular type which can be symmetrical or asymmetrical in shape. The fuzzy logic control with triangular membership function gives better performance than FLC (fuzzy logic control) with trapezoidal under any voltage conditions.

5 63 J.Jeraldin et al., 2016 Fig. 7: Membership function blocks A membership function can have different shapes. The membership functions which are commonly used are the triangular type which can be symmetrical or asymmetrical in shape. The shape of a truncated triangle (symmetrical or unsymmetrical) is the trapezoidal membership function. Fig. 8: Triangular Membership function Triangle(x: a, b, c) = {0 x < a} {(x - a)/ (b - a) a x b} {(c - x)/ (c - b) b x c} {0 x > c} Fig. 9: Trapezoidal Membership function Trapezoidal(x: a, b, c, d) = {0 x < a} {(x - a) / (b - a) a x < b} {1 b x < c} {(d - x) / (d - c) c x d} {0 x d} C. Fuzzy Rules: ANFIS uses hybrid learning rule combines the gradient rule and least square estimate. The linear type membership function has only one output in which each rule is generated separately. To determine each rule s consequent equations, rule extraction method first determines the number of rules and antecedent membership functions which then uses linear least squares estimation. From the FIS editor, the rule editor dialog box appears with IF-THEN rules.

6 64 J.Jeraldin et al., 2016 Table I: True table of membership function ERROR / IN NB NM NS Z PS PM PB ERROR NB NB NB NB NB NM NS Z NM NB NM NM NM NS Z PS NS NB NM NS NS Z PS PS Z NM NM NS Z PS PM PM PS NS NS Z PS PS PM PM PM NS Z PS PM PM PM PB PB Z PS PM PB PB PB PB D. Fuzzy Structure: ANFIS model structure has two inputs as error and change in error and the input membership function are NB,NM,NS,Z,PS,PM and PB. Combination if rules are framed using IF-THEN condition to get the desired output. The logical operations can be AND condition or OR condition. The surface viewer dialog box clearly shows the structure of the error, change in error and the output. Fig. 10: Structural view of ANFIS IV. Simulation Results: The simulation of the proposed system is carried out in MATLAB/ SIMULINK environment. Models can be built using SIMULINK which contains a large number of blocks. In the main SIMULINK window blocks are arranged in block libraries which are accessed. The properties of the blocks in the SIMULINK library can be modified to suit the practical conditions thereby helping to obtain accurate results. Fig 11 shows the simulation graph of input voltage source at three phase voltage source. Fig. 11: Input voltage source Fig 12 shows the simulation graph of input current source at three phase current source.

7 65 J.Jeraldin et al., 2016 Fig. 12: Input current source Fig 13 shows the simulation result obtained in the harmonic distortion analysis. Fig. 13: Output of harmonic distortion Fig 14 shows the simulation results obtained in the harmonic without ripples. Fig. 14: Output of harmonics without ripples Fig 15 shows the simulation graph of voltage and current at varied intervals.

8 66 J.Jeraldin et al., 2016 Fig. 15: Output of voltage and current Fig 16 shows the simulation graph of injected voltage and current where injection is done to reduce the ripples. Fig. 16: Output of injected voltage and current Fig 17 shows the simulation graph of voltage in load at three phases. Fig. 17: Output of voltage in load

9 67 J.Jeraldin et al., 2016 Fig 18 shows the simualation graph of current in load at three phases. Fig. 18: Output of current in load Fig 19 shows the simulation graph of load at three phases. Fig. 19: Output of load Conclusion: This paper presents the modeling and simulation of shunt active power filter controlled by a fuzzy logic controller. This paper also presents a fuzzy logic based three phase shunt active power filter for current harmonic elimination and reactive power compensation. In the proposed system, to improve the power quality of shunt active filter by mitigating the harmonics, the fuzzy logic control with different membership functions is developed. The fuzzy logic control with triangular membership function gives better performance than FLC with trapezoidal membership function under any voltage conditions. Proposed system is capable to maintain the compensated source currents sinusoidal and almost distortion less irrespective of supply voltage and load current conditions by providing compensation without the need of passive filter. The system has the dynamic response of the proposed approach under varying load condition, frequency estimation of 49 to 51Hz and the THD is found below 4% while maintaining power factor operation. The simulation results are performed in MATLAB/SIMULINK. REFERENCES Adzuki Abdul Salam and Nik Azran Ab Hadi., Fuzzy Logic Controller for Shunt Active Power. 4 th International Conference on Engineering Technology and Technopreneuship ICE2T. Bhim Singh., Kamalal-Haddad and Ambrish Chandra., A Review of Active Power Filters for Power Quality Improvement. IEEE Transactions on Industrial Electronics, 46: 5. George Adam., Alina Georgiana Stan (baciu)., Gheorghe Livint., An Adaptive Hysteresis Band Current Control for Three Phase Active Power Filter Using Fuzzy Logic. International Conference and Exposition on Electrical and Power Engineering. Hamad., M.S., Fahmy. and M. Abdel-Geliel, Power Quality Improvement of a Single Phase Grid- Connected PV System with Fuzzy MPPT Controller. IEEE /13.

10 68 J.Jeraldin et al., 2016 Jang., J.-S. R., Fuzzy Modeling Using Generalized Neural Networks and Kalman Filter Algorithm. Proc. of the Ninth National Conf. on Artificial Intelligence (AAAI-91), pp: Jang., J.-S. R., ANFIS: Adaptive-Network-based Fuzzy Inference Systems. IEEE Transactions on Systems, Man, and Cybernetics, 23(3): Jang., J.-S. R. and C.-T. Sun, Neuro-Fuzzy Modeling and Control. Proceedings of the IEEE. Kuo-Ching Tseng and Chi-Chih Huang, High Step Up High Effeciency Interleaved Converter with Voltage Multiplier Module for Renewable Energy System. IEEE Transactions on Industrial Electronics, 61: 3. Maged. F. Naguib and Luiz. A.C. Lopes, Harmonic Reduction in Current Source Converters Using Fuzzy Logic. IEEE Transactions on Power Electronics, 25: 1. Metin Kesler and Engin Ozdemir., Operation of Shunt Active Power Filter under Unbalanced and Distorted Load Conditions. Electrical Education Department, Unuttepe, Turkey. Ngac Ky Nguyen., Djaffar Ould Abdeslam., Patrice Wira., Damien Flieller., Jean Merckle, Artificial Neural Network for Harmonic Currents Identification in Active Power Filtering Schemes. IEEE /08. Nitin Gupta., S.P. Singh and S.P. Dubey, Fuzzy Logic Controlled Shunt Active Power Filter for Reactive Power Compensation and Harmonic Elimination. International Conference on Computer and Communication Technology ICCCT. Nitin Gupta., S.P. Singh and S.P. Dubey, Neural Network Based Shunt Active Power Filter Harmonic and Reactive Power Compensation under Non Ideal Mains Voltage. IEEE /10. Ouamri Bachir and Ahmed Foitih Zoubir., Adaptive Neuro Fuzzy Inference System Based Control of PUMA 600 Robot Manipulator. International Journal of Electrical and Computer Engineering (IJECE) 2: 1. Pablo Acuna., Luis Moran., Marco Rivera., Juan Dixon., Jose Rodriguez., Improved Active Power Filter Performance for Renewable Power Generation Systems. IEEE Transactions on Power Electronics, 29: 2. Musa., S., M.A.M. Radzi., H. Hisham and N.I. Abdul Wahab., Fuzzy Logic Controller Based Three Phase Shunt Active Power Filter for Harmonics Reduction. IEEE /14. Sangu Ravindra., Dr.V.C.Veera Reddy and Dr. S. Sivanagaraju., Design of Shunt Active Power Filter to Eliminate the Harmonic Currents and to Compensate the Reactive Power under Distorted and or Imbalanced Source Voltage in Steady State. International Journal of Engineering Trends and Technology, 2: 3. Suresh Mikkili and Anup Kumar Panda., Mitigation of Harmonics Using Fuzzy Logic Controlled Shunt Active Power Filter with Different Membership Function by Instantaneous Power Theory. IEEE /12. Tzung Pei Hong and Chai Ying lee. Induction of Fuzzy Rules and Membership Function from Training Examples. Institute of Electrical Engineering, 30067, Taiwan. Revised January 1995 and Revised 1 August Wang., L.-X., Adaptive Fuzzy Systems and Control: Design and Stability Analysis, Prentice Hall.