Application of Artificial Neural Network on Diagnosing Location and Extent of Disk Space Variations in Transformer Windings Using Frequency Response Analysis

Abstract

Power transformers are vital components in a power system. One of the most effective and widely accepted ways in identifying mechanical defects of power transformer windings is frequency response analysis (FRA). In this paper, Disk space variations (DSVs), common transformer winding faults, are applied to the 20 kV windings of a 1.6 MVA distribution power transformer in various locations and intensities, and their corresponding frequency responses are computed. It is demonstrated that the designed artificial neural network (ANN) with the proposed novel numerical index called fitting percentage (FP) is capable of diagnosing accurate fault locations and fault extents. The performance of the prepared ANN has been assessed via various classification metrics. The experimental results verify the effectiveness of the proposed method in diagnosing the extent and location of the DSV defects.