Examination on the Denoising Methods for Electrical and Acoustic Emission Partial Discharge Signals in Oil

Partial discharge (PD) measurements either through electrical or acoustic emission approaches can be subjected to noises that arise from different sources. In this study, the examination on the denoising methods for electrical and acoustic emission PD signal is carried out. The PD was produced through needle-plane electrodes configuration. Once the voltage reached to 30 kV, the electrical and acoustic emission PD signals were recorded and additive white Gaussian noise (AWGN) was introduced. These signals were then denoised using moving average (MA), finite impulse response (FIR) low/high-pass filters, and discrete wavelet transform (DWT) methods. The denoising methods were evaluated through ratio to noise level (RNL), normalized root mean square error (NRMSE) and normalized correlation coefficient (NCC). In addition, the computation times for all denoising methods were also recorded. Based on RNL, NRMSE and NCC indexes, the performances of the denoising methods were analyzed through normalization based on the coefficient of variation (𝐶𝑣). Based on the current study, it is found that DWT performs well to denoise the electrical PD signal based on the RNL and NRMSE 𝐶𝑣 index while MA has a good denoising NCC and computation time 𝐶𝑣 index for acoustic emission PD signal.