Recognition of Partial Discharge in Switchgear Based on Kohonen Network

Abstract

Recognition for partial discharge in switchgear was faced with the problem of uncontrollable interference and difficulty of initial parameters determination, so a method based on Kohonen network was presented to improve such problems. By designing defects that meet the characteristics of discharge in switchgear multiple samples were collected, and statistical parameters are extracted from two-dimensional distributions. The influence of Kohonen network's parameters on its recognition effect was investigated, after which the recognition effect is optimized. Then by comparing recognition result of this network and other commonly used recognition algorithms, it is proved that Kohonen network has high stability and good recognition performance when facing switchgear's partial discharge recognition problem.