Neuro-evolutionary genetic algorithm for global MPPT under partial shading conditions: a comparative analysis with PSO

Abstract

Maximizing power extraction from photovoltaic (PV) systems is crucial for their overall efficiency. However, under partial shading conditions (PSCs), the power-voltage curve shows several points of maximum power. This phenomenon often leads to traditional maximum power point tracking (MPPT) algorithms getting stuck at suboptimal local peaks, resulting in substantial energy losses. To solve this, we introduce a novel neuro-evolutionary genetic algorithm (NEGA) for global MPPT. This hybrid algorithm integrates a neural network to intelligently guide the evolutionary search process, improving its GMPP tracking. The performance of the NEGA controller is rigorously compared against the widely used particle swarm optimization (PSO) algorithm via MATLAB/Simulink simulations across various irradiance scenarios. Results under severe PSCs demonstrate NEGA's superior tracking efficiency of 98.69%, far exceeding PSO's 76.02%. Moreover, NEGA achieves a faster convergence time of 0.1 s under dynamic irradiance, compared to 0.6s for PSO. The study concludes that NEGA is a robust and highly efficient solution for global MPPT, ensuring maximum power harvesting from PV systems under challenging operating conditions.