Recognising multiple partial discharge sources in power transformers by wavelet analysis of UHF signals

Partial discharges (PD) in a transformer generate transient UHF signals that resonate within the tank. Where more than one source of PD or interference is present, the ability to separate incoming signals depending on their point of origin would enhance the diagnostic process. A wavelet-based method for recognising UHF signals from PD sources at different positions in the transformer tank, even when the signal is received at only one UHF sensor, is described. Time-frequency characteristics of the UHF PD signals are established using the wavelet transform, providing a means of mapping the energy distribution of the signal. A similarity function is then defined to provide a measure of the similarity between the various wavelet coefficient distributions. The procedure is demonstrated by means of laboratory experiments and application of the technique to data obtained from a transformer in the field.