Efficiency Analysis of Artificial Intelligence and Conventional Maximum Power Point Tracking Methods in Photovoltaic Systems

Abstract

This study investigates the performance of different maximum power point tracking (MPPT) methods in a photovoltaic (PV) energy system, focusing on Artificial Neural Networks (ANNs), reinforcement learning (RL), and conventional MPPT approaches. The primary objective is to evaluate the efficiency of these methods in maximizing PV energy production under varying environmental conditions, while also analyzing their impact on battery state-of-charge (SOC) and load management. The system is modeled using MATLAB, incorporating real-world climate data from Ankara, including monthly solar irradiance, temperature, and sunlight hours. The ANN-based MPPT employs a multi-layer perceptron (MLP) to improve PV efficiency, while the RL-based MPPT utilizes Q-learning to optimize energy production. A conventional MPPT method serves as a baseline for comparison. Simulations are conducted on an hourly and monthly basis, considering a 7.5 kW PV system with a 20 kWh battery system. The results indicate that both ANN and RL methods outperform the conventional MPPT in terms of annual energy production, with RL achieving the highest efficiency gains. Additionally, the ANN and RL methods demonstrate improved battery SOC management, reducing energy losses. The study concludes that advanced MPPT techniques, particularly RL, offer significant potential for enhancing PV system performance, making them viable solutions for renewable energy optimization.

Keywords

• maximum power point tracking
• reinforcement learning
• artificial neural networks
• renewable energy