Methods for comparing frequency response analysis measurements

Abstract

This paper presents and evaluates a number of methods for comparing the results of frequency response analysis (FRA) measurements on power and distribution transformers. It begins with a short description of FRA and its uses. This includes a short review of the literature. Existing methods for comparing the results of FRA measurements are presented and discussed. A case is made for the increased use of statistical indicators. Two different statistical indicators are presented: the correlation coefficient and the spectrum deviation. The mathematical definition of each is given. The two indicators are applied to case studies, based on actual applications of FRA to power and distribution transformers. The performance of the indicators in these case studies is evaluated, considering the importance of avoiding both false positive and false negative results. It is concluded that the correlation coefficient is a useful statistical indicator in most circumstances. It is further concluded that the spectrum deviation is an unsuitable indicator, as it is particularly prone to give false negative results in some circumstances. Recommendations are made concerning the evaluation of other statistical indicators which may be proposed in the future.