Gas Insulated Switchgear (GIS) has been widely used in power system.The most common failure is the partial discharge (PD) resulting frominsulation defects. And different types of PD damage GIS in different ways.Therefore, researching on the fingerprint of insulation defects in GIS andidentifying the type of discharge correctly are of significant value to helpensure the safe and reliable operation, value the insulation condition andmake rational maintenance strategy for GIS.In this paper, the GIS PD mechanism and the various insulation defectmodels in GIS are introduced in detail. The extraction method ofcharacteristic parameters and the reduction algorithm for the feature spaceare described. Based on the discussion, this paper selects8statisticalparameters to describe the typical features of PD. Principal ComponentAnalysis (PCA) method is used to reduce the dimensionality of the featurespace.Four kinds of typical PD models are designed to simulate the actualinsulation defects occurred during GIS operation. A GIS UHF PD systemis constructed to acquire PD signals and3-D plots are showed. Statistical parameters are extracted from the acquired PD signals. A four-typeSupport Vector Machine Classifier is constructed based on Support VectorMachine (SVM) Algorithm, and then the PD type is identified by votingmethod. Experimental results show that the proposed method possesses ahigh recognition rate and can effectively identify these four types of GISPD. This paper concludes the research work done in this subject, and pointout the part needs to be improved in the future. |