An optimized artificial neural network based maximum power point tracking for photovoltaic systems using chicken swarm optimization

Abstract

Conventional Maximum Power Point Tracking (MPPT) methods suffer from low accuracy and slow response under rapidly changing weather and load conditions. This paper proposes an enhanced training technique to improve the performance of conventional Artificial Neural Network (ANN)-based MPPT for photovoltaic (PV) systems. The method employs the Chicken Swarm Optimization (CSO) algorithm to determine the optimal network topology, weights, and bias values. The proposed ANN-based MPPT model takes solar irradiance and temperature as inputs to predict the voltage corresponding to the Maximum Power Point (MPP). This predicted voltage serves as a reference and is compared with the actual PV voltage to generate an error signal. The resulting error is used to regulate the duty cycle of the PWM signal driving the DC-DC boost converter. Meteorological data collected from the Modibbo Adama University weather station were used to ensure accurate ANN training and real-world applicability. A comprehensive PV system model including solar panels, a DC-DC boost converter, MPPT controllers, and varying loads was developed using MATLAB/Simulink to evaluate the proposed CSO-ANN MPPT method. Simulation results under various operating scenarios show that the proposed CSO-based tuning method gave the best ANN network topology with optimum weights and bias values. This significantly improves its power tracking performance compared to the conventional Perturb and Observe (P&O) method. The proposed approach achieved the lowest ripple factor of 10.01%, in contrast to 32.02% obtained with the conventional P&O algorithm.