Detection of Power Transformer Winding Faults using Orthogonal Wavelet Filter Bank

Transformer plays very important role in power system utility. Its protection against various faults is necessary to avoid catastrophic failures. The terminal behaviour of the transformer utters about its health. Thus, analysing the terminal behaviour is helpful in predicting condition of the transformer. In this paper, frequency response analysis (FRA) is deployed to capture terminal behaviour of the winding corresponding to its healthy and faulty states. The acquired FRA signals are supplied to the Daubechies Orthogonal Wavelet Filter Bank and are decomposed into various subbands (SBs). Afterwards, Log Energy (LE) feature is extracted corresponding to each decomposed subband. The extracted features are then classified using decision tree method. The proposed methodology is implemented on the equivalent circuit model of the transformer winding to classify its FRA signals into normal and faulty states. Result shows that the FRA signals are classified properly and accuracy of 98.3% is achieved. The statistical parameters clearly indicate the difference between healthy and faulty signals.