GIS Insulation Fault Diagnosis based on Detection of SF6 decomposition Products

Partial discharge (PD) and partial over-thermal (POT) are the major failure modes of gas insulated switchgear (GIS), the measurement of which are among the most important methods of diagnosing insulation failure in GIS. The decomposition products of SF6 under PD and POT can be used to detect and recognize PD and POT, but feature parameters must be selected first. In this paper, three kinds of PD defect models and a gas chamber were designed to simulate PD. A heating rod and a real disconnector chamber were used to simulate POT. By reviewing the decomposition mechanism of SF6 and analyzing the experiment results, three concentration ratios were chosen as feature parameters, that is, c(SO2F2+SOF2+SO2)/c(CO2), c(SOF2)/c(SO2) and c(SOF2+SO2)/c(SO2F2). The physical significance of the proposed three concentration ratios was also analyzed. Finally, fuzzy c-means clustering was employed to test the effectiveness of using these three concentration ratios as feature parameters to expose the difference between PDs and POT. Results showed that the proposed feature parameters performed well and could successfully recognize different kinds of PD and POT.