Investigation of Partial Discharge on Typical Defects with UHF Detection Method for GIS

Five typical insulation defect models in 110kV SF6-insulated switchgear (GIS) are deigned in this paper. Phase Resolved Partial Discharge (PRPD) spectrums for the five defects are constructed by Ultra High Frequency (UHF) method, and a dataset of 390 samples is established. By analysis of the PD distribution, a feature extraction method based on statistical theory is adopted. Finally, fuzzy c-means clustering (FCM) is introduced to analyze the natural similarity among the various typical insulation defects. The results show that the overall recognition rate is of 90.5% high and most of the defects can be roughly unified to their respective categories, except for some intersections of the surface contamination and viod in spacer defects.