Analysis of Machine Learning Techniques for Electrical Fault Detection in Photovoltaic Panels

Abstract

Because solar energy is sustainable, economical, low-maintenance, and pollution-free, its use has grown lately. However, a number of mechanical and electrical issues with the solar panels restrict the solar farms' performance. In this research, machine learning (ML) methods such as Naïve Bayes (NB), Random Forest (RF), Support Vector Machine (SVM), Decision Tree (DT) and K-Nearest Neighbors (KNN) are used to analyse an autonomous solar panel failure-detection system. Line current values, DC voltage, current range, DC power, radiation, temperature, and defect resistance are among the electrical characteristics used by the classifiers. Further, the synthetic minority oversampling techniques (SMOTE) is applied to diminish the data scarcity issue and class imbalance issue by generating artificial samples. For four-class defect detection, the DT classifier achieved an accuracy of 95%, precision of 95%, recall of 95.83%, and F1-score of 94.94%.For the SMOTE-augmented dataset, the DT provides improved accuracy of 97%, recall of 97.0%, precision of 97%, and F1 Score of 97.01% for four-class classification. ML algorithms demonstrate automatic feature-learning capabilities that accelerate the finding of photovoltaic (PV) panels' electrical faults.