SFA-YOLO: A method for detecting hot spots in photovoltaic modules

Due to complex environmental disturbances and the limitations of detection technologies, the accuracy and reliability of hot spot detection for photovoltaic modules are facing significant challenges. In actual operation, environmental factors such as changes in lighting conditions, shadow occlusion, dust accumulation or snow coverage may lead to misjudgment and missed detection in infrared imaging detection. This paper proposes an improved YOLOv11 network model, which is capable of integrating spatio-temporal features. By defining the Spatio-temporal Feature Aggregation Module (SFA) and combining it with the boundary box regression loss function, the feature and texture information of multiple receiving fields can be supplemented and fused. Finally, a comparative experiment was conducted on the self-collected dataset to verify the effectiveness of the method proposed in this paper.