Substation Equipment Anomaly Detection Method Improved Based on AnoGAN

This paper proposes an improved method of substation equipment defect detection based on the adversarial network to improve the accuracy of substation equipment condition detection. The algorithm uses defect-free data to train the adversarial network generator and uses the axial attention mechanism combined with jump connection to improve the network's understanding of image details. The detection accuracy of this algorithm for 11 types of defect scenes in substations reaches 86.95%, which is 2.89% higher than that of the mainstream YOLO. It can be applied to the defect detection task of substation equipment.