Computer-aided analysis of discharge patterns

Abstract

In recent years investigations of the use of digital evaluation of partial discharges have become increasingly important. Such methods of evaluation greatly improve the recognition of discharge patterns and the correlation of these patterns with various types of defects. Conventional methods of oscillographic observation provide a limited recognition of defects causing discharges. A powerful interpretation of the discharge signals can be achieved when a large number of discharge quantities are measured simultaneously, making use of their stochastic character. A computer-aided discharge analyser has been built for continuous and multiparameter registration. This paper concentrates on the analysis of several parameters for recognizing patterns of partial discharges in a dielectric. A combination of statistical and discharge parameters was studied to discriminate between different types of cavities. Two phase-angle quantities, pulse counts and mean pulse height distributions as a function of phase angle, were processed, using different statistical operators such as skewness, kurtosis and modified cross correlation. The tests on samples with different internal defects have shown that significant differences exist in the behaviour of discharge parameters. Three of these parameters provide a good description of discharge and offer a good discrimination between different internal defects.