Abnormal Scene Image Recognition Method for Intelligent Operation and Maintenance of Electrical Equipment in Substations

The utilization of robot inspections and accurate, rapid identification of fault targets in images through artificial intelligence technology represents an important research area in the field of intelligent operation and maintenance for substations. To address the issue of massive image data collected by robot inspections, an abnormal scene image recognition method is proposed for intelligent operation and maintenance of electrical equipment in substations, with consideration for both the accuracy and real-time performance of object detection. Firstly, the research status and development trend of intelligent operation and maintenance technology of substation electrical equipment is elaborated. Subsequently, a fault detection model for intelligent operation and maintenance of electrical equipment in substations is proposed based on YOLO v5 deep neural network. To improve the detection speed of the model, Cross Stage Partial (CSP) and Spatial Pyramid Pooling Fast (SPPF) structures are utilised to reduce model parameters. Finally, the effectiveness of the proposed model is validated through two application scenarios: substation foreign object suspension image recognition and thermal defect detection based on infrared images. The experimental results demonstrate that the model achieves a detection accuracy of 99.8% in foreign object suspension detection for electrical equipment, with an average frame rate of 149 frames/s. Moreover, the model has an accuracy of 98.9% for the detection of thermal defects in infrared images of electrical equipment, with an average frame rate of 123 frames/s. These results satisfy the application requirements for real-time monitoring.