Clustering of transformer condition using frequency response analysis based on k-means and GOA

Abstract

Frequency response analysis (FRA) is considered as the most popular and reliable method to detect mechanical deformations within power transformers. Despite this popularity, interpretation of FRA signatures has not yet been standardized worldwide. Correct interpretation of FRA signatures results in reliable diagnosis of the transformer mechanical integrity which facilitates timely and proper remedial action. As a further step towards the full understanding of the analysis of FRA signatures, this paper presents a k-means method to cluster power transformers under different fault types. In this regard, series of FRA measurements has been conducted on various transformer models under different fault types. Then, a feature based on interval maximum to global maximum (IMGM) is extracted from the obtained FRA measurements to facilitate data clustering. Using grasshopper optimization algorithm (GOA), the centers of the clusters are determined and by applying the data obtained from operating transformers, the performance of the proposed method is evaluated and compared under different cases.