Mechanical fault is a common fault type in gas insulated switchgear (GIS). The mechanical performance of GIS is very important for the safe and stable operation of power system. In order to achieve accurate diagnosis of GIS mechanical fault, this paper proposes a feature fusion method based on VGG16 and multi-source signals. Firstly, fault simulation test was carried out on a real GIS prototype and feature signals were collected. Wavelet transform was used to obtain the wavelet scale coefficient map of feature signals and then the map was fused. Then adversarial generative network (WGAN) is used to expand the fused samples. Finally, VGG16 network is used to realize sample discrimination, and then complete GIS mechanical fault diagnosis. Experimental results show that the fault diagnosis accuracy of the proposed method is up to 95%, which is higher than that of traditional diagnosis methods, and the fusion samples have richer features than single signals. Meanwhile, the sample set expanded by data enhancement method can effectively solve the problem of insufficient generalization ability of deep learning classifier caused by the lack of samples.

GIS,Mechanical Performance,Artificial Intelligence