Detection of overheating defects in substation disconnectors based on improved YOLOv11

The human inspection method for detecting overheating defects in substation disconnect switches is inefficient and insufficient for meeting the safety management requirements of substation equipment. To address the above issue, we propose an improved YOLOv11-based model for detecting overheating defects in substation disconnect switches. Firstly, the Coordinate Attention mechanism is introduced into the original C3k2 module to enhance the feature extraction capability of the network for finegrained features. secondly, a Multi-scale Feature Fusion network (MFFNet) is designed to replace the original Path Aggregation Network (PANet), which enhances the network's ability to characterise the multi-scale features. Finally, we introduce SIoU to improve the bounding-box regression loss function in the Head, in order to improve the model's detection accuracy and adaptability to complex scenaries. Extensive experiments conducted on an infrared image dataset of substation disconnect switches demonstrate that our improved YOLOv11 achieves a mean Average Precision (mAP) of 90.84%, which is 5.12% higher than the benchmark model. It can be better deployed on mobile terminals for substation safety monitoring.