Substation Equipment Fault Identification Based on UFPN-Fuse Network

To address the problem of poor compatibility between spatial location and information extraction in existing substation equipment fault identification methods based on deep learning, this study proposes a fault identification method based on a UFPN-fuse network. First, the infrared image of the faulty device was segmented using the improved U-Net network, and the fault point features were extracted. Subsequently, the fault features and the original infrared image are fused in the improved FPN-fuse network to strengthen the contour of the fault point in the infrared image. In this way, fault location is achieved by enhancing the visual effect of the image while retaining the detailed information of the fault. Experimental results show that, compared with the comparison algorithms, the proposed algorithm achieves an average increase of 7.83% for SF, 7.48% for MI, 10.62% for AG, and 8.38% for VIF.