Application of neural networks for the control strategy for series active power filters

This paper presents a methodology for the determination of a control reference in an active power filter using multilayer neural networks. The method requires one cycle of distorted voltage waves generated by an usual nonlinear load. One of the major problems in the industry plants is related to the voltage harmonic distortion. The control strategy based on neural networks will be compared to the classical SRF (synchronous reference frame) control algorithm. The validity of such method is proved based upon simulation results and through implementation.