Denoising of partial discharge signal using a hybrid framework of total variation denoising-autoencoder

Abstract

Partial discharge (PD) measurement is an essential task for assessing the condition of electrical insulations. Different types of noises, such as random noise and discrete spectral interference creep into the captured PD signals during their measurement. As a result, there is a need for an efficient denoising technique to reduce the effects of these disturbances during the PD measurement. This paper proposes a denoising scheme through a hybrid approach, which uses a total variation denoising filter followed by a denoising autoencoder. The proposed filter is used here to denoise partial discharge (PD) signal corrupted with artificially induced white Gaussian noise. The efficacy of the proposed work is evaluated with the help of some denoising metrics such as signal-to-noise ratio (SNR), root mean square error (RMSE), peak signal-to-noise ratio (PSNR), and mean absolute error (MAE). The results amply indicate the notable performance of the proposed scheme, compared to the existing state-of-the-art PD denoising techniques.