To address the issues in existing algorithms for processing cable partial discharge (PD) signals

such as oscillations

incomplete denoising

low computational efficiency

and complex key parameter selection

this paper proposes a denoising method based on singular value decomposition (SVD) and Tunable Q-factor Wavelet Transform (TQWT)

referred to as the SVD_TQWT denoising algorithm. First

the noisy PD signal is transformed via Fourier Transform to obtain its spectrum

and SVD is applied to estimate the number of periodic narrowband interferences

followed by constructing a Hankel matrix to remove them. Next

the PD signal in the absence of narrowband interferences is decomposed using TQWT

and separated into high-frequency and low-frequency subbands by sample entropy. Furthermore

the white noise in each subband is removed using the group sparse total variation denoising (GSTVD) algorithm and an improved wavelet threshold denoising (WTD) algorithm

respectively. Thus

pure PD signal is obtained. Denoising results of simulated and measured PD signals show that SVD_TQWT denoising algorithm has significantly improved in a variety of evaluation indicators compared with the other three denoising algorithms

which provides strong supports for subsequent research on cable partial discharge location and pattern recognition.