Partial discharge localisation on power transformers using neural networks combined with sectional winding transfer functions as knowledge base

In this paper, the applicability of different neural classifiers trained by sectional winding transfer functions (SWTFs) of transformer coils for evaluating wideband electrical measured partial discharges (PD) is presented and discussed. Based on PD measurements on a distribution transformer in the laboratory, it is demonstrated that the introduced technique enables determination of the apparent charge and adequate localisation of the PD origin. The performance and the reproducibility of this method leads to a short outlook on planned investigations and the possibilities for integrating this technique in online condition monitoring systems