Three-Phase Active Power Filter Based on Neural Network Model Reference Adaptive Control

Aiming to solve the problem of harmonics and low power factor in power system, a shunt three-phase active power filter (APF) based on neural network model reference adaptive control (NNMRAC) is designed to suppress the harmonics and improve power factor. The APF is based on a three-phase pulse width modulated (PWM) voltage source inverter (VSI) and corresponding control circuit. The DC capacitor voltage and the output current errors between reference model and APF are neural network’s inputs; neural network’s outputs are instruction currents, which are input into three hysteretic current controllers to generate the switching pulses of the VSI. Simulation results using MATLAB verify the effectiveness of the algorithm under different load currents and different power factor.