Denoising of UHF Partial Discharge Signals by Using VMD Based on Shannon Entropy and Kurtosis-Approximation Entropy

Abstract

The ultra-high frequency (UHF) partial discharge (PD) monitoring method can effectively monitor the internal insulation state of the power equipment. When, however, there are some different types of noise in the substation, it's difficult to simultaneously suppress them with the existed methods. Therefore, Two methods are proposed for PD signal denoising in this paper, namely the VMD based on Shannon entropy and the VMD based on kurtosis-approximation entropy (APEN). The variational mode decomposition (VMD) algorithm is firstly used to decompose the original PD signal into several modes. Then, the Shannon entropy value and kurtosis-approximation entropy value are calculated respectively. Next, the Shannon-based rule and the APEN-based rule are designed to identify the effective components. Finally, the signal is reconstructed by adding the selected modes. Experiments were carried out under mixed noise interference of different intensities. Furthermore, output signal-to-noise ratio (SNR), root mean squared error (RMSE), signal distortion ratio (SDR), and noise rejection ratio (NRR) were also used as indexes to evaluate the denoising performance of the algorithms. Denoising results based on these indexes for simulative and experimental data show that the proposed two methods can filter various noises on a large scale. Compared with EMD, EEMD and CMSE-based VMD, the two methods can obviously perform better, and the latter has the best denoising effect.