Application of ANFIS Controlled Shunt Active Filter for Harmonic Reduction Authors : Chun-Tang Chao, Chi-Jo Wang, Cheng-Ting Hsu, Nguyen ThiHoaiNam Presented by : Nguyen ThiHoai Nam
OUTLINE • Introduction • Shunt Active Filter Modeling • Control System Design • Simulation Results • Conclusions
1. Introduction • Reason choosing this research topic • Reduction harmonic method • Proposed controller: ANFIS (Adaptive Neuro Fuzzy Inference System)
2. Shunt Active Filter Modeling • Active filter is a power electronic device based on the use of inverters • Shunt Active Power Filter is connected in a common point connection between the source of power system and the load system which present the source of the polluting currents circulating in the power system lines Fig. 1. Power system with non-linear load and shunt active filter.
2. Shunt Active Filter Modeling (1) (2) Fig. 1. Power system with non-linear load and shunt active filter. (3) From (1), (2) we have: Formula (3) indicates that purpose of shunt active power filter is intended to generate exactly the same harmonics contained in the polluting current iLbut with opposite phase.
2. Shunt Active Filter Modeling The mathematical model can be extracted from the single-phase equivalent scheme by Fig. 2. Fig. 2. Single-phase equivalent scheme (7) (4) (5) (8) (6)
3. Control System Design3.1 Control structure of Active Filter Fig. 3 is applied to control AF producing current track with the load current harmonic Fig. 3. The active filter control structure. Where: AF is active filter; BPF is band pass filter; LPF is low pass filter; PWM is pulse width modulation.
3. Control System Design3.1 Control structure of Active Filter Fig. 4. The simulation model of electrical power system with active filter.
3. Control System Design3.1 Control structure of Active Filter Fig. 6. Active Filter structure using IGBTs Fig. 5. The controller structure.
3. Control System Design3.2 Fuzzy Logic Controller for AF • Fig. 7. Fuzzy controller synoptic diagram Fig. 8 Rule viewer window. Fig. 9 Relationship between e, de, u
3. Control System Design3.3 ANFIS Architecture for AF • Jang originally presented the Adaptive Neuro-Fuzzy Inference System technique in 1993 . Jang combined both Fuzzy Logic and Neural Network to produce a powerful processing tool named Neuro-Fuzzy Systems that have both Neural Network and Fuzzy Logic advantages and the most common one is ANFIS. Actually, this tool is like a fuzzy inference system, but the difference is in the use of a back propagation algorithm for minimizing the error.
3. Control System Design3.3 ANFIS Architecture for AF Fig. 10. ANFIS architecture • Layer 1 consists of input variables • Layer 2 is membership layer • Layer 3 is rule layer • Layer 4 is defuzzificationlayer • Layer 5 is output layer
3. Control System Design3.3 ANFIS Architecture for AF Fig. 11. 500 patterns are loaded into the ANFIS editor tool Fig. 12. Result of the ANFIS model testing with training data
3. Control System Design3.3 ANFIS Architecture for AF Fig. 13. Membership functions of input e. Fig. 14 Tuned membership functions of input e
3. Control System Design3.3 ANFIS Architecture for AF Fig. 15. Membership functions of input de. Fig. 16 Tuned membership functions of input de
4. Simulation Results Table 1. Simulation parameters
4. Simulation Results Fig. 17. Supply current isa waveform before applying the AF. Fig. 18. Harmonic spectrum of isabefore applying AF
4. Simulation Results • Fig. 19. Supply current isawaveform after applying AF using FLC • Fig. 20. Harmonic spectrum of isa after applying AF using FLC
4. Simulation Results • Fig. 21. Supply current isa waveform after applying AF using ANFIS • Fig. 22. Harmonic spectrum of isa after applying AF using ANFIS
4. Simulation Results Fig. 23. AF current and its reference with ANFIS.
4. Simulation Results Table 2. Total Harmonic Distortion (THD) (%) in different running conditions of load
5. Conclusions In this work, the FLC and ANFIS are developed to reduce the harmonic current for nonlinear loads through running simulation in Matlab/Simulink environment. Importantly, the applied ANFIS controller is better than the fuzzy controller and can also be used to improve the control performance of nonlinear systems. Experimental results and simulations show that the resulting shunt active filter presents good dynamic and steady-state response.Harmonic pollution is always kept under IEEE 519standards.
Thank you for listening