A probabilistic feature to determine type and extent of winding mechanical defects in power transformers

Abstract

Frequency response analysis (FRA) is a widely-used method to detect axial displacement (AD) and radial deformation (RD) in windings of power transformers, because of its high sensitivity to small amount of mechanical defects. Interpretation of frequency response curves has been the most intricate problem of FRA method. To solve it, different numerical indices have been introduced by researchers to evaluate the frequency response of power transformers, but (1) the researchers have not discussed on genesis origin of the proposed indices and (2) most of these indices can not present a regular and linear behavior. In this paper, a probabilistic feature has been utilized to demonstrate how an efficient index originates, to diagnosis axial displacement and radial deformation in windings of power transformers. The resulted index presents a nearly linear mapping between frequency response curves and their related defect extent. To verify usefulness of this index, it has been applied to several sets of measured frequency response curves. It is illustrated that the extracted index is able to determine extent of the defects. To discriminate between axial displacement and radial deformation, local behavior of TF curves has been used.