Identification of simultaneously occurring dynamic disc-to-disc insulation failures in transformer winding under impulse excitation

A method for identification and localization of isolated as well as simultaneous dynamic series insulation failures in transformer winding has been proposed. Trained wavelet network has been employed to identify the fault characteristics, viz. location(s), nature(s) and type of unknown impulse insulation failures. The network identifies the simultaneous dynamic insulation failures using the cross-wavelet transform based features extracted from recorded winding currents resulting impulse excitation. Four different fault emulators have been used to emulate different combinations of simultaneous dynamic disc-to-disc insulation failures in transformer winding. It has been found that the developed network with extracted features has identified the fault characteristics of simultaneous dynamic series insulation failures of transformer winding insulation within ±9% of winding length with good accuracy. The characteristics as well as experimentation details of simultaneously occurring dynamic disc-to-disc insulation failure(s), cross-wavelet transform based feature extraction and wavelet based fault characteristics identification are explained.