A measuring system based on modern signal processing methods for partial discharge recognition and localization on-site

The paper describes a modular structured measuring system for partial Discharge (PD) recognition and localization on-site. The appearing noises are suppressed digitally with different filter algorithms each being best adapted to the respective noise signals. The suppression of sinusoidal continuous noises is realized by means of digital filters, whereas synchronous pulse shaped noises are eliminated by using adaptive correlation algorithms. A separation of the PD from noise pulses with stochastical appearance is performed with pattern recognition algorithms. Investigations of pulses measured on transformers show, that the simultaneous employment of several algorithms leads to an increase in the reliability of the evaluation. Also, a monitoring operation without any a priori knowledge about appearing signal shapes becomes possible. Besides the noise suppression, the same pattern recognition algorithms with changed parameters make a contribution to the localization of PD faults.