Signal processing and study of the ripple influence in PD patterns for measurements under HVDC stress

摘要

The traditional phase resolved partial discharge (PRPD) patterns used for partial discharges (PD) sources recognition when alternating voltages are applied is not applicable for direct voltage stress. The phase displacement of the PD pulse regarding the zero crossings of the AC voltage signal has been traditionally used for distinguishing different types of insulation defects (internal cavity, surface and corona effect). However, under HVDC stress no zero crossing exits and consequently PRPD patterns cannot be used for PD source identification. In the last years, it has been demonstrated that the PD patterns in DC can be correlated through the time interval and the amplitude difference between consecutive PD pulses. The problem appears when more than one single PD source is involved in the insulation and also when significant electromagnetic interference is present. When two or more pulse sources appear in the discrete insulation system (e.g. in a cable terminal), two consecutive PD pulses can be originated by the different pulse sources and the PD pattern can lead to wrong insulation diagnosis. An efficient way for on-line PD monitoring of HVDC systems is to remove the background noise. After filtering, the background noise is removed and each pulse can be detected and automatically placed with a better accuracy along the cable system where it is generated; the localization is performed determining the time delay of the pulse arrival times to different PD sensors. This paper presents how different PD sources placed in the same insulation site can be separated by means of a 3D PD clustering on the basis of the analysis of the PD wave-shape. The paper also analyzes the dependency of PD patterns obtained under HVDC stress with respect to the HVDC level and the ripple voltage level.