Incipient fault diagnosis in power transformers by clustering and adapted KNN

Abstract

Dissolved Gas Analysis (DGA) is one of the proven methods for incipient fault diagnosis in power transformers. In this paper, a novel DGA method is proposed based on a clustering and cumulative voting technique to resolve the conflicts that take place in the Duval Triangles, Rogers' Ratios and IEC Ratios Method. Clustering technique groups the highly similar faults into a cluster and makes a virtual boundary between dissimilar data. The k-Nearest Neighbor (KNN) algorithm is used for indexing the three nearest neighbors from an unknown transformer data point and allows them to vote for single or multiple faults categories. The cumulative votes have been used to identify a transformer fault category. Performances of the proposed method have been compared with different established methods. The experimental classifications with both published and utility provided data show that the proposed method can significantly improve the incipient fault diagnosis accuracy in power transformers.