Artificial Neural Network Based Shunt Active Power Filter for Power Quality Improvement

Abstract

The increase in non-linear loads in the electrical distribution system has resulted in harmonics, which affect the power system’s stability and performance. The project focuses on mitigating the current harmonics of the system by implementing a Shunt Active Power Filter (SAPF), optimized using an Artificial Neural Network (ANN). The proposed system ensures the generation of compensating current through the Shunt Active Power Filter (SAPF), which helps in mitigating the harmonics present in the system. The project showcases the gradual improvement in power quality using different techniques for optimization in the conventional Shunt Active Power Filter (SAPF). This paper presents a comparative analysis of harmonic reduction techniques for a non-linear load system. The uncompensated system shows a Total Harmonic Distortion (THD) of 28.34%. Applying a conventional Shunt Active Power Filter (SAPF), PI-tuned SAPF, and PSO-optimized PI-SAPF reduces THD to 14.01%, 3.43%, and 1.64%, respectively. An Artificial Neural Network (ANN) further enhances the PSO-PI controller through adaptive real-time optimization. MATLAB/Simulink simulations demonstrate the proposed ANN-based SAPF achieves a THD of 1.41%, offering superior harmonic suppression compared to traditional methods. This demonstrates the significant improvement in the power quality by effectively mitigating the harmonics present in the system.