Automized Recognition of Partial Discharges in Cavities

Abstract

A simultaneous analysis of several parameters was used to study the possibilities of recognizing specific patterns of partial discharges in a dielectric. A computer aided discharge analyzer was used to test two types of cavities: dielectric bounded and electrode bounded. The following discharge parameters were registered: ignition voltage, discharge magnitude and phase position. The discharge data were processed using statistical tools like cross-correlation and skewness. A combination of statistical and discharge parameters offers a strong discriminator between the two types of cavities studied and can be a usefull step into developing an expert system for the recognition of different kinds of defects in insulation.