The increasing use of power electronic inverters in electrical drives is leading to the unexpected inception of partial discharges (PD) within the insulation system of electric machine stator windings. Unfortunately, the short rise and fall times of such voltages, below 100 nanoseconds in most cases, create high-magnitude conducted and radiated interferences that affect the measurements. Detecting partial discharges in these systems with inductive or electromagnetic sensors turns to be an important challenge, especially when interference and partial discharges share the same bandwidth. In this work, the authors propose a technique to detect the occurrence of PD in such systems, even when they are masked by disturbances. The proposed technique combines an analog high-pass filter stage and a digital band-pass to select a proper spectral interval. The resulting signal is converted to the base-band, where two metrics based on the peak-to-average power ratio are proposed. Experimental results obtained with the signals acquired with two antennas show that both proposed metrics exhibit a significant increase, about ten times greater, when the applied voltage exceeds the PD inception voltage. These contributions define new markers for the detection of PD events created by pulsed voltages with overlapped overshoots with promising results.
electric motors; electric vehicles; electrical insulation; inverters; partial discharges; radio-frequency sensors; spectral analysis