Dual-sensor correlation for wideband HFCT partial discharge detection and interference suppression in transformers

Abstract

On-site partial discharge (PD) detection of power transformers is challenged by strong electromagnetic interferences, including converter pulsation, corona, and power-side harmonics, whose spectra overlap with PD signals. A dual-channel synchronous detection system based on self-developed wideband high-frequency current transformer (HFCT) sensors was developed. The HFCT had an operating band of 3–50 MHz and an average transfer impedance of 16 mV/mA, whereas the signal-conditioning system covered 150 kHz–60 MHz. A 14-bit acquisition unit sampled both channels at 250 MSps using a common clock, corresponding to a sampling interval of 4 ns. Satellite timing synchronized distributed acquisition units with an accuracy better than 100 ns. A dual-sensor correlation discrimination method was proposed using time-of-arrival difference, polarity sign, and amplitude transfer ratio as weighted decision parameters in the propagation-path domain. Experiments on a 220 kV/2 MVA single-phase transformer simulation platform using a dataset of 2,800 frames, comprising 2,400 defect–interference combination frames, 240 background-noise frames, and 160 calibration-signal frames, achieved 91.3% recognition accuracy, 14.2 dB SNR improvement, 37.8 ms processing time per 4,096-sample event window, and a 2.4% detection blind zone. The event-level evaluation included 800 PD events and 1,600 interference events extracted from the defect–interference combination frames. Compared with wavelet thresholding, EEMD, and adaptive SVD, the proposed method improved accuracy by 3.7–8.6 percentage points, with recalculated electrical-localization errors ranging from 0.21 to 0.40 m, while the minimum identifiable apparent charge for air-gap discharge was approximately 5.8 pC. The method provides an effective solution for reliable on-site PD detection under complex electromagnetic environments.