Application of Classifiers for On-line Monitoring of Transformer Winding Axial Displacement by Electromagnetic Non-destructive Testing

Abstract

Transformer winding on-line monitoring using electromagnetic non-destructive testing has been suggested in this article. As a test object, a simplified model of transformer has been used. The winding axial displacement can be modeled on the test object. The scattering parameters of the test object can be measured and stored in a database. To detect the axial displacement, two indices have been defined using the magnitude and phase of scattering parameters. The k-nearest neighbor and decision tree classifiers have been used for the detection of the winding axial displacement and its value. The accuracy of the k-nearest neighbor method to find the axial displacement value has been improved by using a proposed k-nearest neighbor regression algorithm. The comparison of the average error of two classifiers shows the superiority of the k-nearest neighbor regression over the decision tree classifier.