Comparison of IEC 60599 gas ratios and an integrated fuzzy-evidential reasoning approach in fault identification using dissolved gas analysis

Abstract

Fault identification in oil-filled electrical equipment is a task that requires accurate methods of data collection and analysis. When the oil-paper insulation system in an electrical equipment is subjected to excessive operating stresses various gases are formed which are characteristic of the stresses generating them. The most reliable method of detecting incipient faults due to these stresses in such equipment is the dissolved gas analysis (DGA). Various DGA tools are available for fault identification and they mainly fall into two categories, numerical and artificial intelligence methods. This paper presents the comparison of a numerical method; IEC 60599 gas ratios and an artificial intelligence method; fuzzy-evidential reasoning in fault identification. After careful analysis of various faults documented in IEC and IEEE literature, and classifying them into five categories based on IEC 60599 standards, it is possible to describe them using fuzzy trapezoidal membership functions. By aggregating the resultant partial membership outputs using fuzzy logic or evidential reasoning, faults can be identified. One hundred and seventeen faults cases documented in IEC TC 10 databases are used to compare the effectiveness of the fault identification using these techniques. The fuzzy logic and fuzzy-evidential reasoning techniques appear to yield more accurate results than IEC 60599 gas ratio method. This work also evaluates the degree of normalcy for an equipment considered fault free.