Recognition of defects in high voltage cables using statistical tools

The analysis of partial discharges (PD) using conventional discharge detection (bandwidth about 400 kHz) is presented. It was found that the shape of the phase-position distribution contains a large amount of information on the origin and the type of the discharge. These shapes are therefore analyzed, using statistical operators like skewness, kurtosis, and cross-correlation factor. It is known that statistical operators are characteristic for actual discharge sources, and the combination of 15 statistical operators can form the fingerprint of a discharge. To use these fingerprints for the recognition of unknown defects, an algorithm called `recognition rate´ was developed. The algorithm was used to reveal the origin of discharges in high-voltage cable. The results show that the origin of typical discharges in a cable can be traced back in this way