Abstract: Infrared detection of substation electrical equipment is challenging owing to remote imaging conditions and small-target sizes. To address these challenges, this study proposes an improved YOLOv8-based detection method. In the improved YOLOv8 model, the Coordinate Attention (CA) mechanism module was introduced to strengthen feature extraction and feature fusion capabilities of the network model. The original model loss function, CIoU, was optimized to SIoU to reduce misjudgments of the model, and the detection head, YOLO-Head, was supplemented to improve detection of remote and small targets. With these three improvements, the average detection accuracy, expressed as mean Average Precision (mAP), on infrared images of substation electrical equipment increased from 89.28% to 93.57%, and the comprehensive index of precision-recall rate (P-R) also increased by 3.2%.