De-noising of partial discharge signal using wavelet transform for GIS under HVDC

Detection and analysis of partial discharge (PD) have been regard as the most effective method for condition monitoring and asset management of gas-insulated structures (GIS) in the power system. However, PD detection sensitivity and accuracy are greatly influenced by on-site noise and interference, resulting in failures in PD severity assessment, defect identification or localization. Although de-noising of PD signal under AC was well studied, related investigations under DC have not been carried out. With the rapid development of HVDC technology, it is a new challenge to eliminate noise from PD signal under DC for diagnosis of related power facilities. Therefore, this paper dealt with the discrimination of PD signal based on wavelet transform (WT) techniques for HVDC GIS, aiming to improve the sensitivity and accuracy of insulation diagnosis. Experimental setup was configured to generate PD signal under DC and four types of artificial defects were fabricated to simulate typical insulation defects in GIS. The WT techniques were used to discriminate PD pulse sequences from background noise, amplitude modulation radio interference, non-sinusoidal noise, and switching impulse and the effectiveness was compared with a high-pass filter.