Using Machine Learning for Detection and Classification of Single and Mixed Faults in Photovoltaic Systems

Abstract

This study proposes a machine learning-based approach for the detection and classification of single and mixed faults in a photovoltaic system based on representative electrical parameters. Six classification algorithms were compared in order to identify the most effective model. In order to reduce the impact of class imbalance and improve the quality of the training data, the SMOTE-Tomek technique was applied. Stratified cross-validation was also used to evaluate the stability and generalization capacity of the models. The results show that the XGBoost algorithm offers the best performance with an accuracy of 90.79% in cross-validation and 92.26% on the test set. In addition, this model proved particularly effective in distinguishing between classes with very similar electrical signatures, particularly in the case of mixed faults. These results confirm the relevance of boosting techniques for reliable and accurate diagnosis of photovoltaic systems.