The partial discharge caused by insulation defects inside the electric power equipment contains the characteristic information which can be used for the evaluation of insulation state. The characteristic information can effectively identify the types of different partial discharges. The existing BP neural network or SVDD pattern recognition method is very independent in the parameter selection. However

the characteristic quantities of different discharge types are overlapped and non-linear in the distribution

the BP neural network is easy to fall into the local optimum and the recognition rate is not high

SVDD in free metal particles defect recognition effect is not good. Based on the algorithm of AP clustering and GPAM-PSO

a SA-SVDD algorithm for local discharge pattern recognition of solid insulated switchgear is proposed. To solve the shortcomings of the traditional pattern recognition algorithm in the parameter selection and training methods

by training the classifier under different discharge types to achieve accurate identification of different discharge types. The simulation results show that the method can identify different PD types

and the recognition rate and convergence rate are greatly improved compared with the traditional methods so that the electric power personnel can judge the partial discharge type accurately and form the corresponding repair scheme.