The internal insulation defects in gas-insulated switchgear (GIS) can be reflected by partial discharge, so it is significant to recognize the type of partial discharge (PD) in GIS correctly. Four kinds of typical defection of the GIS are designed on the GIS intensive care research system. The extraction of the characteristics of PD gray image is the recognition features based on the PD sample data of the four kinds defects under different voltage levels. At the same time, considering the effect of interference on the partial discharge signal, GK fuzzy clustering algorithm is used to further process fractal feature and to extract the analysis of characteristics. At last, the PD type recognition device is designed based on the LS-SVC. Experimental results show that using the proposed method the PD type within GIS can be correctly recognized. In addition, the method proposed is stable and possesses higher recognition rate than the artificial neural network method.