Partial Discharge Signal Denoising Based on Block Thresholding

In high voltage engineering, it is of great importance to monitor partial discharge (PD) of critical apparatus. On-site PD signal is usually contaminated by white noise. Therefore, it is necessary to denoise the measured signal to acquire pure PD. The most widely used method is the traditional wavelet thresholding. In this paper, a modified version of the traditional wavelet thresholding is proposed, i.e. block thresholding is used to replace regular thresholding. Wavelet coefficients are divided into blocks which will be regarded as a whole part, then each block is thresholded. Several common thresholding functions are compared with block thresholding, and the computation results show that block thresholding is better than traditional wavelet thresholding.