Evaluation of statistical interpretation methods for frequency response analysis based winding fault detection of transformers

Abstract

Frequency Response Analysis (FRA) is an offline condition monitoring technique commonly used to diagnose the mechanical integrity of transformers. Due to high accuracy and sensitivity, FRA has become a common practice in the industry for assessment of winding health. However, interpretation of FRA measurements is still limited to analysis by field engineers relying entirely on their past experiences and expertise. In order to aid the FRA based Condition monitoring, this paper evaluates different statistical indicators, developing and collecting them on a single platform, namely an automated interpretation package for FRA signatures. A MATLAB based GUI has been developed to aid detection of transformer faults. Experimental verification consisting of inter-turn short circuit fault studies on a 15 kVA cast-resin transformer to examine developed automated package has also been reported as part of this work.