Rapid assessment of photovoltaic module health using an improved YOLOv11

Photovoltaic (PV) energy systems, recognized for their environmental sustainability and renewable characteristics, have achieved broad implementation and exhibit substantial potential for further development. The operational condition of PV modules critically influences both power generation efficiency and output quality, necessitating timely condition assessment to enable early fault identification and maintenance interventions. This investigation employs image data acquired by a gimbal-mounted camera system onboard an unmanned aerial vehicle, processed through a Jetson Orin NX edge computing platform. The YOLOv11 algorithm is enhanced through several technical advancements. Implementation of a regional attention mechanism, multi-channel feature fusion architecture, optimized loss function formulation, and an improved fuzzy K-means clustering strategy. These modifications collectively enable more precise detection of hot-spot defects in PV modules. The developed evaluation framework is subsequently applied to assess the module's health status. Experimental results indicate that the enhanced YOLO algorithm achieves superior fault recognition accuracy compared to previous versions, demonstrating robust performance across diverse background complexities and illumination conditions. The proposed health assessment methodology proves both reliable and effective in operational evaluations.