A Denoising Method of Partial Discharge Signals Employing Wavelet Kernel-Aided Deep Learning Framework

Noises can contaminate acoustic partial discharge (PD) signals, resulting in inaccurate detection and interpretation. As a result, denoising acoustic PD signals has achieved significant attention to enable precise diagnosis and analysis. This manuscript presents a wavelet kernel (WK) aided deep learning (DL) framework for denoising acoustic PD signals obtained from a high-voltage power apparatus. For this work, the measured signals are contaminated with low-level random noises and then fed to a fully convolutional network (FCN) based autoencoder (AE) to perform denoising operations. Unlike the classical FCN, the proposed FCN employs mother wavelets as convolutional kernels for processing PD data for necessary operation. Both qualitative and quantitative assessments demonstrate the effectiveness of the proposed method when compared to alternative denoising techniques. Furthermore, the validation of the proposed method extends to two variations of simulated PD signals, each subject to varying levels of white Gaussian noise. The proposed scheme can also monitor and diagnose other signal modalities in real-time.