Time-Frequency Hybrid Suppression Method for Discrete Spectral Interference in Partial Discharge Signal

To solve the serious waveform distortion and high computational complexity of existing methods for discrete spectral interference (DSI) suppression in partial discharge (PD), this paper presents a time-frequency hybrid method based on mathematical morphology filter (MMF) and spectrum correction. The method includes DSI frequencies identification in frequency domain and DSI waveforms reconstruction in time domain. Firstly, logarithmic amplitude spectrum (LAS) of noisy signal is filtered by a MMF, which then be subtracted from original LAS; secondly, the number and coarse frequency values of DSI are identified by adopting the classical threshold criterion from subtracted LAS; thirdly, the spectrum correction method is introduced to estimate the accurate frequency, amplitude and phase of aforementioned DSI, which then can be used to reconstruct DSI waveforms; at last, the DSI is subtracted from the noisy signal to achieve the goal of DSI suppression. Simulation data show that, compared to optimal harmonic wavelet packet transform based method and blind source separation based methods, the proposed method is full-automatic, and it possesses a high interference suppression performance and a high waveform fidelity, thus it is suit for PD online measurement.