Pattern classification of partial discharge in high voltage equipment by regression analysis

This paper introduces a statistical classification in regression model to classify partial discharge (PD) patterns into four categories in corona: high voltage side in air, corona at low voltage side in air, surface in air, and internal discharge. There are nine independent variables from fingerprint analysis which mainly are skewness, kurtosis, asymmetry and cross correlation following Phi - q - n PD patterns. The variable independent data are divided into two group: the forms for training the model and the pattern for test the model. The created algorithm investigates the best of parameter group from all of independence variables for creating the regression model. In this work, the group of five best parameters for classification are selected. The results show, that only five parameters in the model has a good performance to classify PD pattern with the classification accuracy of 100%.