Analysis of Levenberg Marquardt - ANN based reference current generation for control of shunt active power filter

Abstract

Shunt active power filter (SAPF) is an inevitable solution that needs to be installed by industrial consumers to prevent the power quality issues due to the non-sinusoidal currents drawn by non-linear loads. SAPF provides harmonic current compensation and reactive power management to eliminate the current related power quality issues. Various reference current generation schemes such as d-q theory, instantaneous reactive power theory, discrete Fourier transform, PI controller, fuzzy logic etc. are reported. Artificial neural network (ANN) is an emerging concept in SAPF for reference current generation. In this paper, detailed analysis of Levenberg Marquardt–feedforward ANN based reference current generation scheme for the control of SAPF is presented. This paper mainly focuses on successful implementation of Levenberg Marquardt algorithm for training the feedforward ANN network for reference current generation with different non linear and unbalanced load condition. Performance analysis of Leven Marquardt – ANN based reference current generation for the control of SAPF under different operating condition is carried out in MATLAB/SIMULINK platform.