Diagnosis of power transformers through frequency response analysis by poles and zeros shifts identification

Detection of abnormalities within transformers is a field of study aimed at preventing future permanent damage. Current automatic diagnosis methodologies include some proposals that use the Sweep Frequency Response Analysis (SFRA) technique as a condition assessment criterion. SFRA has a high sensitivity to various failures in windings and iron core. However, these methodologies are not completely robust due to the parameters used for discrimination, generally statistical indicators. This paper shows a new diagnosis methodology, which has been successfully applied in other fields, based on Vector Fitting analysis for poles and zero shifts recognition. Vector Fitting is a suitable tool as a condition assessment in SFRA traces analysis, where the analysis is done using poles and zeros location of the transfer function. As poles and zeros are represented in complex domain, the Principal Component Analysis (PCA) is introduced for data complex transformation to the real domain with the same information.