Diagnosis and clustering of power transformer winding fault types by cross‐correlation and clustering analysis of FRA results

Abstract

The power transformer is one of the vital and substantial elements of each country's power grid which not only require high investment, but they are also important in terms of economy, social, political, and strategy. Since this equipment is exposed to different electrical and mechanical winding faults during operation, they should be monitored continuously. One of the main monitoring methods is the use of frequency response analysis (FRA), which has a high sensitivity. The main challenge of the FRA is that the detecting task of the status of the transformer is done by a specialist and with a visual evaluation of the records. To overcome this problem, first, frequency responses in the healthy and present states are calculated through simulation of electrical and mechanical fault in the winding of the transformer and then, new statistical methods are used to interpret FRA results based on the obtained transfer function. In this study, for the first time, clustering analysis and cross-correlation methods are used to interpret FRA results for clustering and diagnosis of different short circuits turns, axial displacement, and radial deformation. Results and simulations verify ability and advantage of these methods in detection and determination of different faults.