Differential Protection of Three Phase Transformer Using Artificial Neural Network

This paper presents the use of an Artificial Neural Network (ANN) for differential protection of three phase transformer, which makes the discrimination between the internal faults and the other disturbance like magnetizing inrush and over fluxing. The discrete wavelet transform has been used for feature extraction and certain statistical parameters of wavelet decomposition at various levels acts as an input for ANN. The authors have used two layer feed forward neural network with only one hidden layer for classification. The proposed scheme provides good results than conventional harmonic restrain scheme with overall classification accuracy of 97.8 %.