As a characterization of the phenomenon of high voltage equipment insulation status

discriminate for the type of partial discharge has the vital significance for switchgear insulation condition assessment. The different partial discharge types have some common characteristics

at the same time there are individual differences. The more comprehensive descriptions describe the characteristics of partial discharge

the higher the accuracy of the determination.This paper studies the different types of partial discharge in time domain

frequency domain and statistical mappingdifference

put forwards the characteristic parameters

which are able to more fully describe the types of partial discharge.Considering the dimension of feature

proposes a fault type identification method for BP neural network and SVM basedon RST. The studies result that compared with BP neural network

the SVM method has stronger ability of identificationand greatly less time on the design period of classifier and the training. Through the combination of SVM and RST

themethod can not only improve the identification ability of the standard partial discharge type

also identify more untypical partial discharge signals by joining the new case study.