SK-FRCNN: A Fault Detection Method for Hot Spots on Photovoltaic Panels

Summary: As solar capacity installed worldwide continues to grow, advanced inspection systems are becoming increasingly important. In this study, a novel automatic multi-stage model is proposed to detect panel defects on aerial images captured by unmanned aerial vehicles using YOLOv3 network and computer vision techniques. The model demonstrates versatility in processing different types of images and detecting various defects, providing a more efficient data-driven maintenance strategy. The model's performance has been validated in two large PV plants in Italy.