Repetitive learning control with fuzzy tuning for three-phase shunt active power filters in stationary d-q frame

Active power filter (APF) has been widely recognized as a substitute for classical passive L-C filter in eliminating the power line harmonics/reactive current due to nonlinear loads and switching devices. Under nonlinear loads, for eliminating harmonic currents with APF, the inverter must be commanded to output compensating currents which are time- varying with rapid vibrations. Therefore, the tracking capability of the current controller of the inverter dominates the performance of the APF. The purpose of this paper is the development of a simple and efficient control scheme for the current tracking control of shunt APFs. In the stationary frame, the iterative learning control scheme is adopted to command the APF generate desired compensating currents. In particular, the use of a simple proportional-type (P-type) current-cycle learning mechanism proves to be adequate for the controller to learn the desired current waveform iteratively. The learning gain is then online tuned through a fuzzy inference engine to improve convergent rate and tracking accuracy in practice. Extensive simulation and experimental results are presented to validate the tracking performance of the proposed method. Comparative study with the conventional PI controller is also given.