The invention discloses a GIS partial discharge fault detection method based on a characteristic ultrahigh frequency signal. The method comprises the steps that one, a typical insulation defect fault characteristic sample library based on the characteristic ultrahigh frequency signal is established; and two, discharge fault type identification is performed on GIS partial discharge with the unknown defect. Self-adaptive calibration is performed on the ultrahigh frequency signal so as to confirm the characteristic frequency bands of various typical defects. Discharge data of the characteristic frequency bands are processed so as to acquire discharge characteristic and perform mode identification. A new technical support is provided for GIS partial discharge mode identification by the method so that the result of GIS partial discharge fault detection is enabled to be more accurate and the new idea and method are provided for GIS partial discharge mode identification.