Current Harmonic Compensation by Shunt Active Power Filter Based on Adaptive Linear Neural Networks Using Symmetrical Components

Abstract

Generation of reference current constitutes an important part in the control of active power filter, so a new neural identification scheme for the current harmonic extraction is proposed, that is, the fundamental positive and negative sequence components of non-linear load current are extracted directly by adaptive linear neural filter to generate reference current under non-ideal conditions. And a phase-locked loop system is used to the track the actual grid frequency so the adaptive linear neural filter adapts with grid frequency fluctuations. The reference currents are then introduced to the control scheme based on fuzzy logic variable hysteresis band current control, so the active power filter can track the reference current in a fast dynamic response. The proposed scheme is studied analytically and tested by simulation using Matlab/Simulink. The validity and effectiveness of the harmonic compensation are confirmed by the simulation results.