Identification of Transformer Winding Deformation Fault Type Based on Signal Distance and SVM

Abstract

Listen

Translate article

Aiming at the problem that it is easy to cause misjudgment by the manual experience to diagnose the winding deformation type in frequency response method, this paper proposed a new transformer winding deformation fault type identification method based on signal distance and support vector machine (SVM). The method used the mutual distance index and combined with the nonlinear classification performance of SVM to diagnose the degree of winding deformation and identify the winding deformation fault type. The sub-frequency band mutual distance, correlation coefficient and the resonance frequency amplitude change rate were selected as the feature quantity of SVM. And the grid search method and cross-validation were used as parameter optimization method of the SVM. The simulation experiments show that the selection of the sub-frequency band mutual distance as the feature quantity can reach the accuracy of 96.4286%, better than correlation coefficient 89.2857% and the frequency-amplitude change rate of the resonance point 92.8517%. So the proposed sub-frequency band mutual distance as the SVM feature quantity is suitable for the transformer winding deformation fault type identification, and the mutual distance is a good fault feature for identifying transformer winding faults.