Challenges of ANN in Adapting to Abrupt Parameters Variations for Shunt Active Filter

Abstract

The purpose of the shunt active power filter (SAPF) goes beyond simply improving power factor; it also aims to mitigate unwanted harmonic currents generated by nonlinear loads. This research introduces a SAPF equipped with an identification and control method utilizing artificial neural networks (ANN). Various techniques are employed for harmonic identification, including traditional p-q theory and the relatively recent approach of artificial neural networks. Achieving satisfactory identification and control characteristics is challenging when dealing with the nonlinearity of the system, particularly in situations involving rapid nonlinear load fluctuations. This study represents a systematic effort to address this challenge by enhancing the SAPF with a method for harmonics identification and DC link voltage control based on ANN. This approach is tested on the SAPF under conditions of fast, nonlinear load variations. The study includes both computer simulations and experimental results, which provide evidence of the viability and effectiveness of the proposed active power filter.