A novel three-phase active filter based on neural networks and sliding mode control

Abstract

In this paper, a new three-phase active power filter (APF) is proposed. This filter can be successfully utilised for harmonic mitigation and reactive power compensation in electric power systems supplying nonlinear loads. An adaptive linear neural network is adopted for harmonic current detection and reference signal estimation. In addition, a sliding mode technique is used for maintaining sufficient tracking of reference. It controls the generation of PWM gating patterns required for the proper operation of a 3 phase voltage source converter. The APF is shunt connected to the point of common coupling (PCC) to supply the appropriate compensating currents. The integrity of the proposed scheme is fulfilled through digital simulation results. Moreover, the superiority of its performance is proved by comparison to other active filtering approaches. The results show an appreciable improvement in the source current waveforms and a near unity power factor operation is achieved. The presented control system of the APF exhibits robustness and a high dynamic response. The resultant enhancement in the system power quality is expressed by a unified index.