A Fuzzy Logic-based Methodology for Dissolved Gas Analysis to Enhance the Accuracy of Power Transformer Incipient Faults Detection

Abstract

Detecting incipient faults within electrical power transformers at an early stage prevents troubles and guarantees service continuity. Dissolved gas analysis is widely used in evaluating the condition of transformers and detecting incipient faults. The methods that rely on dissolved gases in fault detection could be classified into two types: traditional methods and artificial intelligence-based methods. Traditional methods are frequently and widely used because of their ease and simplicity. The most well-known conventional techniques used are Rogers’ ratio, Doernenburg ratio, Duval triangle-I, and IEC standard code 60599. Despite the spread of these methods and their ease of use, they have drawbacks. The most important of these drawbacks is the overlapping between the various incipient faults and their inability to detect a large number of faulty cases. This relies on codes on which these methods are based. These codes change very sharply that does not occur with the dissolved gases inside the transformer. Accordingly, this paper presents a development for the three main traditional methods based on fuzzy logic intelligent method. The use of Fuzzy logic reduces the sharp changes that occur in the codes by using a suitable membership function for each input and output. The accuracy of the developed methods is examined through a database collected from transformers affiliated to the Egyptian Electricity Holding Company. These transformers have different rates and different service durations. The results comparisons demonstrate that the developed methods are accurate in identifying faults and minimizing the overlapping between faults and cases of unpredictability.