Backpropagation and Particle Swarm Optimization Based Neural Network for Power Quality Improvement in Grid Connected Systems

Abstract

This paper compares back propagation and particle swarm optimization (PSO) in neural network parameter tuning for power quality improvement in distribution grid connected systems. A non-linear load is connected to the distribution system that injects harmonic currents into the grid side. Neural networks are trained using PSO and Backpropagation. Each neural network is then used in a STATCOM controller that is used for power factor correction, reactive power compensation, power balancing, etc. The algorithms are then compared on various aspects, including harmonic content in the resulting signal, weight convergence, settling time, rise time, and deviation of DC link voltage from a constant value. The model is also tested for dynamic loading by applying additional load to the power system. MATLAB 2018a is used to train and test the artificial neural networks for power quality improvement in grid connected systems.