Partial discharge signal recognition method based on singular value decomposition and probabilistic neural network

In the monitoring of electrical equipment, the partial discharge signal type can be identified quickly and accurately, which will be conducive to the safe and stable operation of power grid. In this paper, a partial discharge signal recognition method based on singular value decomposition and probabilistic neural networks(PNN) is proposed. Firstly, from the theory and simulation, it is proved that in the time of partial discharge, the Hankel matrix of partial discharge signal is constructed for singular value decomposition, and fewer singular values can be used to describe the waveform characteristics of PD signal. Then, the effective order of singular value selection of partial discharge is determined by singular value difference spectrum. The singular value energy percentage of the previous singular value is selected as the eigenvector, and the probabilistic neural network optimized by genetic algorithms(GA) is used to identify the source of cable partial discharge signal. The results show that this method can well identify four partial discharge signals: cable body discharge, cable terminal discharge, corona discharge and surface discharge in switchgear.