Transformer abnormal heat accurate identification method based on AHIPDNet

Abstract

Abstract Transformer is an important part of the power system, and its operation status is related to the safety and stability of the power system. In order to effectively improve the recognition accuracy of abnormal transformer heat, an improved YOLOv8 target detection algorithm model is proposed. The introduction of the SPD-Conv convolutional layer effectively improves the model’s recognition ability of the small target, and the introduction of the HAT hybrid attention mechanism and Dyhead multiple attention even if the information focusing on the abnormal hot spot is realized. The HAT hybrid attention mechanism and Dyhead multi-attention mechanism are introduced to better focusing of the model on the information of abnormal hot spots. Finally, the AHIPDNet transformer abnormal heat recognition method is formed. After experimental verification, the algorithm model achieves 93.1% recognition accuracy, which satisfies the demand of accurate recognition and localization of transformer’s abnormal heat in the field.

Keywords

• Abnormal heat detection
• Transformer
• YOLOv8
• Complex background detection