The influence of data size in statistical analysis of power transformer frequency response

Abstract

Frequency Response Analysis (FRA) has been acknowledge as the most reliable and sensitive diagnostic tool to assess mechanical condition of power transformer. During assessing the integrity of transformer, proper and correct comparing interpretation of FRA response is become crucial. This paper presents and evaluates the influence of data size in statistical analysis when comparing the results of transformer frequency response. Three different statistical indicators are presented: the correlation coefficient (CC), the absolute sum of logarithmic error (ASLE) and the comparative standard deviation (CSD). The performance of these three selected statistical indicators while the data size varies is evaluated on simulated response. The same analysis is then repeated on frequency responses which were measured from an actual power transformer to test the validity of the earlier finding. In addition, the method of relative factor that proposed by the Chinese Standard DL 911/2004 is also evaluated and presented in this paper. Overall, this study concluded that the data size of the response does affect the statistical indicators. Recommendation is made that a standard data size is to be established and included in future FRA standard.