Knowledge based interference signal rejection and partial discharge identification from multi-PD sources for condition monitoring of cable systems

Autonomous partial discharge (PD) identification is not only key to the success of, but also one of the major challenges in, PD based condition monitoring of cable systems. In industrial applications, two challenges restrict the implementation of autonomous PD identification. The first challenge is that interference signals are of different shapes and the frequency bandwidths of some interference signals are similar to that of PD signals. The second challenge is that, as PD signals originate from different sources, their characteristics also vary. Sources of PD include cable and cable termination, switch-gear enclosures, motors connected at the end of the cable being monitored, etc. The ability of identifying PDs from the multiplicity of PD sources is therefore in great demand. To overcome the above two challenges, this paper demonstrates a knowledge based method for interference signals rejection and autonomous PD signal identification. The fingerprints of typical interference signals and PD signals emanating from various sources are defined and differentiated in the time domain. Based on the fingerprints, a decision tree based recognition method and a knowledge base has been developed. The method was based on PD on-site testing data collected from more than 300 cables in one power system. It was then applied to another two systems and proved to be effective.