Infrared Image Recognition and Classification of Typical Electrical Equipment in Substation Based on YOLOv5

Infrared images of substations can effectively identify potential faults in electrical equipment. However, due to the fact that infrared images of substations are unstructured data, identifying the information in the images requires the professional experience of a large number of personnel, which not only wastes manpower but also reduces the accuracy of equipment fault diagnosis based on infrared image information. This article first established a typical infrared image dataset of substation equipment, and then optimized the learning rate in the YOLO v5 algorithm model using dynamic attenuation method. The weight attenuation coefficient of the model was optimized to avoid overfitting problems in the dataset; Finally, the recognition and classification processing of typical substation equipment in infrared images were achieved through two model training sessions. The results show that the recognition rate of isolation switches and transformer bushings in infrared images is 92.45%, while the recognition rates of CT, PT, lightning arrester, and circuit breaker are 93.10%, 87.50%, 93.55%, and 90.91%, respectively. The research results of this article not only solve the problem of identifying and accurately classifying electrical equipment in infrared images of substations, but also provide a technical foundation for intelligent diagnosis of equipment faults based on infrared images of substations.