Separation of Spectral Features in Multi-Source PD Measurements Using 3CFRD Method

Power System components are susceptible to insulation failures due to electrical & mechanical stresses, extreme environmental and operating conditions, material ageing, chemical degradation, and poor maintenance practices. The partial discharge (PD) phenomenon is a precursor to insulation degradation and needs to be continuously monitored to avoid catastrophic failures resulting in significant downtime and financial losses. A very important aspect of PD monitoring is noise mitigation and PD source identification. This paper presents a systematic approach for de-noising raw PD measurements using the 3-centre frequency relation diagram (3CFRD) technique and then separating PD sources using phase-resolved partial discharge (PRPD) representations. An experimental setup has been established to acquire PD measurements using IEC-60270 and HFCT methods to capture diversified PD data on surface discharges in XLPE insulation samples combined with corona discharges. The measured PD signals are processed using the 3CFRD method. By a proper refinement of the PD threshold, it was possible to mitigate noise and, further, an optimal selection of suitable centre frequencies at a fixed bandwidth enabled clustering of the PD events in a star diagram. The PRPD plots corresponding to the identified clusters were reconstructed and showed clear features characterising each type of discharge. Such an analysis, combined with relevant field experience, can be used effectively for condition monitoring and health assessment of HV equipment.