Correlation analysis of transformer condition evaluation indexes based on association rules

Abstract

Health state assessment of transformer can help power utility to improve the management level and efficiency. As it is difficult to achieve accurate state evaluation of transformer by using a single indicator, current research focuses on how to build a comprehensive condition evaluation model based on multiple attributes. However, the correlation between multiple attributes does not attract enough attention. Based on the association rules, this paper analyzes the correlation between different indexes for condition monitoring of transformer. The health state of transformer is separately rated based on each single index according to the preset intervals. Afterwards, Apriori algorithm is used to find out the frequent item sets with respect to different indexes and ratings, and the correlation of indexes is judged according to the consistency of evaluation results. The correlation between 7 indexes is investigated by using the real monitoring data of 30 transformers.