Partial Discharge Monitoring Of Dielectrics Using Machine Oriented Algorithmic Techniques

Partial Discharge monitoring provides a tool for the assessment of failure risk for high voltage plant and equipment. The measurement process is complicated by the presence of interference, transients and other sources of disturbance on the system. In practical monitoring systems these disturbances almost certainly fall outside the boundaries of any expert system for PD recognition. Because of the complexities of PD measurement, only a sub-set of the available parameters is considered. There is currently no universal system capable of assessing and analysing partial discharge activity in samples outside of laboratory conditions. This paper describes continuing work to develop machine oriented and self training neural network systems to identify partial discharge taking into account localised characteristics and changing dielectric trends