Adaptive matched filter bank for noise reduction in online partial discharge monitoring

For specific circuits the accuracy of online partial discharge (PD) detection and location (PD-OL) systems is impaired by complications in the cable circuit configuration, such as ring-main-units (RMUs) along the cable under test. For maximum sensitivity matched filtering is applied for extracting PD pulses from the recorded signal. The matched filters are based on a series of prior online system identification measurements and a standard cable model. Especially, with an RMU along the cable, the matched filters only approximate the real waveforms due to signal distortion. This results in sub-optimal detection sensitivity. This paper describes automated procedure for updating the matched filter bank based on measured (large) PD signals. The algorithm analyzes detected pulses to judge whether the current templates perform optimally. If not, a new template is created increasing the PD detection sensitivity and the location accuracy, and also improving the recognition of recurring disturbances. The method is tested on signals measured on a live cable connection.