A Genetic Algorithm Design Method for A Current Controller Employing "Two Ahead" Prediction

This paper introduces an improved predictive control for the control of the current in a three-phase active shunt power filter (ASF). The proposed control system compensates for unavoidable digital controller delays, and provides an accurate prediction of the ASF behaviour in the next sampling period. A "two ahead" current reference prediction is proposed which estimates the filter current two samples in advance. The predictor is based on a second order extrapolation method, and can also account for sudden changes in the reference waveform. A genetic algorithm is used for parameter optimisation in the reference prediction/extrapolation formula. In addition to good tracking performance, the proposed controller is insensitive to measurement noise. The effectiveness of the proposed method is verified through experiment.