An Improved Power Transformer Diagnosis System for Incipient Fault Based on Fuzzy Rough Set Theory

Abstract

Based on fuzzy rough set theory (FRS), the paper is meant to present a new diagnosis system with gas ratios method for transformer incipient fault diagnosis, which not only has the capability of coping with incomplete information inputting and rules reduction just as in conventional rough set theory (RS), but also has the capability of fuzzification of thresholds of continuous values of attributes. Since strict thresholds setting is said to undergo the diagnosis effectiveness, A method of fuzzy subsets extraction from database based on KDD technology is employed to the setting attributes, which improve the power of prevailing IEC/IEEE criteria in fields of strict thresholds setting. Finally, results of testing the proposed diagnosis system on actual dissolved gas records are addressed, which confirmed that rules represented by fuzzy terms and extracted based on FRS allow diagnosis results to be satisfied, and diagnosis system proposed can provide a satisfactory accuracy.