A novel wavelet selection scheme for partial discharge signal denoising

Jiajia Liu, W.H. Siew, John J. Soraghan, Euan A. Morris

Research output: Contribution to conferencePosterpeer-review

64 Downloads (Pure)


Over the past two decades, wavelet-based techniques have been widely used to extract partial discharge (PD) signals from noisy signals. To effectively select the correct technique to minimize the effect of noise on PD detection, three aspects are considered: wavelet selection, decomposition scale, and noise or threshold estimation. For wavelet selection, popular techniques, including correlation-based wavelet selection scheme (CBWSS) and energy-based wavelet selection scheme (EBWSS), are applied to select an appropriate wavelet basis function. These two schemes, however, have their limitations. CBWSS is not as effective as expected when the signal to noise ratio (SNR) is very low. EBWSS selects the optimal wavelet that can maximize the energy ratio of the PD signal in approximation coefficients through wavelet decomposition. It is not strictly true for damped oscillating PD signals, particularly when the decomposition scale increases. As such, a novel wavelet selection scheme, wavelet entropy-based wavelet selection scheme ( WEBWSS), is proposed to provide an alternative to CBWSS and EBWSS for PD denoising. PD signals are simulated and also obtained through laboratory experiments to demonstrate that this new method has better performance in the removal of noise, particularly when SNR is low.
Original languageEnglish
Number of pages4
Publication statusPublished - 29 Nov 2018
EventConference on Electrical Insulation and Dielectric Phenomena - Cancun, Cancun, Mexico
Duration: 21 Oct 201824 Oct 2018


ConferenceConference on Electrical Insulation and Dielectric Phenomena


  • wavelet-based technique
  • partial discharge
  • detection
  • denoising
  • wavelet selection
  • SNR
  • wavelet entropy


Dive into the research topics of 'A novel wavelet selection scheme for partial discharge signal denoising'. Together they form a unique fingerprint.

Cite this