New approach in partial discharge diagnosis and pattern recognition

Digital measuring techniques and diagnosis systems have been developed to assist the test engineer. The most common diagnostic tools are two- or three-dimensional distributions or histograms which exhibit the typical patterns of the partial discharge (PD) activity. The main disadvantage of all these graphs in PD diagnosis is the averaging effect of the statistics used. The authors show a graph of partial discharge activity without any averaging effect, which gives an overview of every single PD pulse of the whole measurement. In this diagram the parameters charge, phase, time and test voltage are included, so that it is possible to summarise all significant information on one computer printout. By seeing PD evolution from period to period, visual recognition and expert classification becomes practicable and the capabilities of an automated personal computer based PD-fault type recognition system is shown. Results of acoustical PD classification are presented and the influence of test voltage on electrical and acoustical PD-recognition is discussed. Additionally, a powerful method to localise the compartment with PD-source in gas-insulated switchgear (GIS) is derived