Comparative Analysis of Machine Learning Algorithms for Fault Detection and Localization in Photovoltaic Systems

This paper presents a comprehensive comparative analysis of four machine learning algorithms (SVM, DT, FNN, RF) for fault detection and localization in photovoltaic (PV) systems. The study aims to address the limitations of traditional methods, such as manual inspections and rule-based approaches, which can be time-consuming, error-prone, and inflexible. By leveraging advanced machine learning techniques, we aim to improve the accuracy, efficiency, and scalability of fault detection and localization in PV systems. We conducted a rigorous evaluation of these algorithms using a unified dataset and performance metrics, focusing on common faults like shading, open circuit, and short circuit. Our findings demonstrate the superior performance of Random Forest (RF) for both fault classification and localization tasks, highlighting its potential for enhancing the reliability and efficiency of PV systems.