Acoustic partial discharge signal denoising using power spectral subtraction

Measuring partial discharge (PD) phenomena in power transformers is often conducted by acoustic emission (AE) method. However, many interference sources are usually encountered with the captured PD signals which negatively affect the PD detection and classification. Thus, an effective and efficient denoising technique is required to suppress such environmental noises. Most denoising attempts aim to address additive white Gaussian noise, which is considered the main ambient interference source coupling with PD signals through data acquisition process. In this paper, we propose a power spectral subtraction denoising (PSSD) method and examine its denoising performance in the presence of modest and severe noise levels. The simulation results verify that PSSD has superior denoising performance when compared to one of the conventional wavelet shrinkage denoising methods. Four evaluation metrics are utilized to confirm the superiority of PSSD: signal-to-noise ratio, root mean square error, cross-correlation coefficient, and reduction in noise level.