Photovoltaic Fault Detection Using I-V Curve Analysis Coupled with Artificial Intelligence Methods

Abstract

Fault diagnosis in photovoltaic (PV) systems is essential for ensuring reliable operation and maximizing energy yield. This paper presents an intelligent PV fault diagnosis framework based on real-time current-voltage (I-V) curve analysis and machine learning techniques. A monitoring device employing a DC/DC buck-boost converter was developed to acquire and display real-time I-V characteristics, enabling continuous condition monitoring and fault detection at the panel level. The acquired I-V data were utilized to classify seven PV operating conditions using two machine learning models: K-Nearest Neighbors (KNN) and AdaBoost. Experimental results demonstrated that AdaBoost achieved a classification accuracy of 95.96%, outperforming KNN, which achieved 93.94%. Furthermore, benchmarking comparisons with representative methods reported in the literature confirmed the superiority of the proposed AdaBoost-based approach. The obtained results highlight the effectiveness, reliability, and practical applicability of integrating machine learning with I-V curve analysis for real-time photovoltaic fault diagnosis and condition monitoring.