Separation of Partial Discharges at DC Voltage using PD Pulse Parameters

The correct identification of partial discharges (PD) and the subsequent risk assessment of the identified defect are of crucial importance and key aspects of a reliable PD monitoring (PDM) system. However, if more than one PD defect or noise pulses occur at the same time, the measurement data will be distorted. Consequently, clustering and separation of PD must be performed prior to classification. At AC voltage, the detection of overlapping defects is easier due to the reference point (the phase position from 0° to 360°), although special methods for separation are also required here. At DC voltage, there is no such reference point, which makes it even more difficult to recognise such cases. The sequence of detected pulses appears more like a random data stream, while the relationship between pulses is unintuitive. This contribution deals with the separation of PD sources at DC voltage as well as the differentiation of noise pulses. PD measurements of typical defects in gas-insulated systems form the basis for the investigations. PD pulse parameters were extracted from the measured pulses. Subsequently, their usability to separate defects from each other at DC voltage was analysed. The results were validated by means of an automatic clustering algorithm.