Dielectric Loss Factor and Circuit Impedance Under Partial Discharge Via Excess Current Detection

Partial discharge measurement plays an important role in the assessment of insulation status and fault diagnosis of power equipment. Presently, the most commonly used industrial applications of PD detections are based on the broadband measurement systems that count amplified PD pulses. The limitations of these traditional measurements are that the system requires a relatively high sampling rate and calculation capability, and is very sensitive to the ambient noise. To eliminate noise interference and save the cost of PD detection system, a technique measurement of excess currents, which measures the dielectric response currents through test object before and after PD activities to capture the activities of PD, has been reported in some recent studies. This work aims to explore the parameters that characterize the PD activities via excess current technology and to seek new ideas for PD detection. In this work, a measurement setup for measurement of PD excess currents averaged from 300 periods is applied and three typical insulation defect models were tested under 50Hz voltages. Excess currents for each case are measured. Then the corresponding dielectric loss factors and circuit impedances are obtained by comparing them to the synchronous voltages. The results show that the loss factors increase after PD while the circuit impedances decrease. Meanwhile, the loss factors continue to increase and the circuit impedances continue to decrease as the voltages turn up where the rate of the change shows more rapid in the needle-plate model. Moreover, the loss factors and circuit impedances in the dielectric barrier model and air void model are respectively higher and lower than the values before PD when the voltages drop below the PD extinction voltages, which does not appear in the needle-plate model. These characteristics provide a new idea for PD detection by monitoring the loss factors or circuit impedances via excess current detection.