Enhancing MPPT in partially shaded PV modules: a novel approach using adaptive reinforcement learning with neural network architecture

Abstract

The occurrence of partial shading in solar power systems presents a substantial challenge with widespread implications, sparking extensive research, notably in the field of maximum power point tracking (MPPT). This study emphasizes the critical process of accurately tracking the maximum power points with the characteristic curves of photovoltaic (PV) modules under real-time, diverse partial shading patterns. It explores the various stages of the tracking process and the methodologies employed for optimization. While conventional methods show effectiveness, they often fall short in swiftly and accurately tracking maximum power points with minimal errors. To address this limitation, this research introduces a novel machine learning approach known as adaptive reinforcement learning with neural network architecture (ARL-NNA) for MPPT. The results obtained from ARL-NNA are compared with existing algorithms using the same experimental data. Furthermore, the outcomes are validated through different factors and processing time measurements. The findings conclusively demonstrate the efficacy and superiority of the proposed algorithm in effectively tracking maximum power points in PV characteristic curves, providing a promising solution for optimizing solar energy generation in partial shading patterns. This study significantly impacts various realms of electrical engineering including power engineering, power electronics, industrial electronics, solid-state electronics, energy technology, and other related field of engineering and technology.