Features of recognising the type of defect in oil-filled equipment using the nomogram method

Abstract

The paper presents the results of analysis of nomograms (graphical areas) of defects in oil-filled equipment of electric networks. The input data were the results of dissolved gas analysis for 2997 transformers and shunt reactors with different types of defects, that is, the training sample. In order to level out possible contradictions in using different criteria, the training sample was split into separate data sets, not only with the same defect type, but also with identical values of characteristic gas ratios, gas percentages and gas-to-gas ratios with maximum concentration. To account for drift of coordinate values of individual nomograms in the obtained arrays, it is proposed to represent defect nomograms in the form of reference regions. The maximum and minimum coordinate values (ratios of each of the gases to the gas with the maximum concentration) obtained for homogeneous arrays of DGA results are used as values of the boundaries of reference regions. The centre of the graphic area coincides with the reference nomogram. As a result, 115 nomograms characteristic of thermal type defects, electrical discharges as well as overheating with different hot spot temperature accompanied by discharges with different energy density and discharges with different energy density accompanied by overheating with different hot spot temperature have been drawn. A brief analysis of the obtained graphic areas is given, the most characteristic damages corresponding to one or another graphic area are considered, and the values of characteristic gas ratios corresponding to the analysed areas are analysed. In the process of analysis it was established that the highest reliability value of defect type recognition can be achieved by simultaneous use of all three diagnostic criteria, namely, values of gas ratios, gas percentages and nomograms (graphic areas) of defects. The obtained results make it possible to significantly increase the number of reference nomograms, which will significantly increase the number of recognisable defects and consequently reduce the risk of accidental damage to oil-filled equipment due to missing defects caused by failure to recognise them.