Optimizing Defect Detection in Photovoltaic Systems Using Machine Learning Based Approaches

For the renewable energy systems, it is imperative to ensure the efficiency and sustainability of the renewable energy system. Photovoltaic (PV) systems are a major contributor in the renewable energy systems. This paper addresses the role of machine learning in optimizing the defect detection in the PV modules using electroluminescence (EL) images. Firstly, the features are extracted using histogram of gradients (HOG) as it is a robust method for capturing the spatial and structural patterns from EL images. Then we investigate the effectiveness of feature selection methods using particle swarm optimization (PSO), genetic algorithm and Bayesian optimization to enhance the defect detection accuracy. The selected features are used to classify defective and non-defective PV cells using the machine learning algorithms namely support vector machines (SVM) and random forest classifiers. The performance of each optimization algorithm is assessed using performance metrics such as accuracy, precision, recall and F1-score. By optimizing feature selection and improving defect detection, this study contributes to enhance the operational efficiency of PV systems. The PSO method for optimized feature selection combined with SVM for defect detection outperformed the other two optimization algorithms giving an accuracy of 85% and precision of 91% on the public EL images dataset used in this work. This work aligns with the global efforts to promote sustainable energy by reducing the losses and improving the maintenance strategies and provides insights for integrating intelligent defect detection frameworks into renewable energy systems.