Application of Fuzzy Rough Set Theory to Power Transformer Faults Diagnosis

This paper is meant to present a new diagnosis measure with gas ratios method for transformer incipient fault. Based on fuzzy rough set (FRS) theory,an information decision system is built,in which some problems in the process of system building are coped with by data mining technology. Firstly,since strict thresholds setting is said to be undergoing the diagnosis effectiveness,continuous attributes are transformed and described based on fuzzy set theory,where the knowledge discovery in database (KDD) technology is used to extract the implied information on fuzzy clustering so as to determine the fuzzy values of attributes and thus the parameters of membership function. Secondly,according to inclusion degree defined in FRS,the formed fuzzy rules are reduced and pruned,where a data-mining algorithm is developed to extract fuzzy rough rules and thus determine the topology of multi-table decision base according to attributes set. Finally,results of testing the proposed diagnosis system on actual dissolved gas records are addressed,which confirms that extracted rules allow diagnosis results to be satisfied with a satisfactory accuracy for diagnosis ratio.