Histogram-based thresholding in discrete wavelet transform for partial discharge signal denoising

Abstract

White noise is a major interference source that affects the partial discharge (PD) signal detection and recognition. Wavelet shrinkage denoising methods can efficiently reject the white noise embedded in the PD signal acquisition and measurement processes. The wavelet threshold determination is a key factor in the quality of noise suppression from signals. A novel threshold estimation technique, namely histogram-based threshold estimation (HBTE), is introduced to obtain the optimal level-dependent wavelet thresholds of noisy partial discharge signals. Unlike existing wavelet thresholding techniques, HBTE obtains two different threshold values for each wavelet subband. The proposed method is applied on measured PD signals at different noise levels. Experimental results show that the proposed thresholding approach outperforms the conventional threshold selection rules in terms of signal-to-noise ratio, cross correlation coefficient, root mean square error, and reduction in noise level.