An improved dissolved gas analysis technique for incipient fault detection

Abstract

Diagnosis of incipient faults is essential for proper maintenance of power transformers. This diagnosis is carried out using established Dissolved Gas Analysis techniques. These techniques include empirical methods for classification of the incipient faults. However, their accuracy is hampered by unavoidable cases of mixture of faults and unresolved diagnosis. This paper intends to overcome these issues using statistical techniques. The regression analysis for different fault scenarios is carried out wherein linear relationship of different fault gases is explored. Firstly the regression equations for fault gas contents in both known and unknown fault cases are obtained. Then the corresponding correlation coefficients for fault gases of each fault type are compared, which can provide an insight for separating and identifying faults in unresolved diagnostic cases. This not only ensures reliability of power supply at customer level but also can improve the economic dividends of power system.