Deep learning in defects detection of PV modules: A review

Abstract

Identifying defects in photovoltaic (PV) modules is essential for ensuring optimal performance and prolonging their operational lifespan. Traditional manual inspection methods are time-consuming, labor-intensive, and subject to human error, necessitating the development of automated, efficient detection techniques. With the increasing scale of PV power plants, there is a pressing need for automated, accurate, and efficient defect detection methods. This review explores the application of deep learning (DL) methods, particularly convolutional neural networks (CNNs), in the identification and classification of PV module defects. Var- ious imaging techniques, including electroluminescence (EL), thermal, and visible spectrum imaging, are discussed for their roles in data acquisition. The importance of preprocessing steps such as image normalization, registration, and segmentation is emphasized to enhance detection accuracy. The review highlights the effectiveness of DL models like MobileNet, VGG-16, and YOLO, and techniques such as transfer learning and data augmentation in improving model performance. Despite achieving high accuracy, challenges such as the need for large datasets and model generalization across different PV modules and environmental conditions remain. The integration of DL with aerial inspection technologies and advance- ments in image processing holds promise for further enhancing the reliability and efficiency of solar energy systems.