Application of signal processing techniques to on-line partial discharge detection in cables

Abstract

Partial discharge measurements have been widely recognized as the most effective diagnostic method for on-line condition assessment. However on-line partial discharge (PD) measurements taken on site are frequently effected by external electrical-magnetic (EM) interference. Extracting low level PD signal from noisy background is a major challenge for on-line condition monitoring. This paper describes on-line PD measurements made on paper-insulated 11 KV cables in an old industrial/suburban substation in Sydney. Ten cables in the substation have been monitored for PD activity on a number of occasions using high frequency current transformers as the sensors. Complementing the PD assessment has been a program of laboratory investigation to simulate PD characteristics useful for the processing procedure. Digital filtering, frequency spectrum analysis, PD waveform cross-correlation and wavelet transforms have been applied to de-noise polluted PD signal from external interference. This paper will present results of partial discharge activity in the cables extracted from the recorded signals on the ten cables in the substation using the de-noising techniques described above. The sensing methods used for signal acquisition will be outlined and the various signal processing techniques used and the results obtained for PD characterization in the cables will be discussed and compared with particular reference to their potential for permanent monitoring applications.