Detection and Classification of Shading and Short-Circuit Faults in a Photovoltaic System Using SVM, DT, and RF with Principal Component Analysis

Abstract

This paper presents a novel approach for the detection and classification of shading and short-circuit faults in photovoltaic (PV) systems by comparing the performance of Support Vector Machines (SVM), Decision Trees (DT), and Random Forests (RF), coupled with Principal Component Analysis (PCA). The study aims to enhance the performance of PV systems by addressing common faults that significantly affect their efficiency and lifespan. A simulated PV system generates fault data, including partial shading and short-circuit under varying environmental conditions. SVM, DT, and RF are applied for fault classification, while PCA is employed for dimensionality reduction to improve computational efficiency. The results show that the DT-PCA and RF-PCA models achieve high classification accuracy, outperforming traditional and machine learning models, with DT-PCA demonstrating the shortest training time with and without PCA. The study highlights the effectiveness of PCA in reducing training time without compromising the accuracy of fault detection. These findings suggest that the proposed hybrid methodology can be deployed for real-time fault detection and diagnosis in PV systems, offering a robust solution for maintaining the reliability and efficiency of renewable energy systems.