Based on the advantages of strong anti-interference and high detection sensitivity

Ultra-High Frequency（UHF）method becomes one of the most effective means on partial discharge（PD）detection. In this paper

according to the common types of PD

the models of suspended discharge

internal discharge and needle plate discharge were designed

and the further research of PD detection based on UHF was implemented. The research showed that

according to the characteristics of PRPD（phased resolved partial discharge）spectra of various PD and their differences

when the characteristic variables of the PRPD spectra was extracting by the way of support vector machine（SVM）combined with logistic regression methods

the types of PD can be identified effectively.