Identification of Partial Discharge State of Transformer Based on Cov-Iradon Transformation-CNN

As an important part of power system, the discharge phenomenon caused by insulation deterioration of power transformer may cause equipment damage. In this paper, corona discharge, air gap discharge and suspension discharge are studied experimentally. Firstly, Covariance (Cov) is used to highlight signal period information and reduce signal noise. Secondly, the inverse Radon transform (Iradon) is used to convert one-dimensional data into two-dimensional images to highlight the characteristic difference between different discharge signals. Finally, convolutional neural network (CNN) is used for state recognition, and the accuracy rate reaches 96.03%.