A multi-domain sparse representation K-singular value decomposition for partial discharge signal denoising under severe noise contamination in gas-insulated switchgear

In gas-insulated switchgear (GIS), noise suppression of partial discharge (PD) signals is essential for guaranteeing the accuracy of insulation condition assessment. However, PD measurements are often severely contaminated by strong background noise in practical GIS monitoring. Developing effective denoising techniques capable of reliably recovering PD waveforms under such harsh noise conditions remains a significant challenge. To address the aforementioned challenges, this study proposes a novel multi-domain sparse representation K-singular value decomposition (MD-KSVD) method for PD signal denoising under severe noise contamination. The method adopts a progressive two-stage denoising strategy, where time-domain K-SVD suppresses narrowband interference and partial Gaussian noise, followed by wavelet-domain K-SVD to further eliminate residual Gaussian noise while preserving transient PD characteristics. The proposed framework achieves robust denoising performance under extremely low signal-to-noise ratio conditions. Extensive experiment validations conducted under varying signal-to-noise ratio circumstances demonstrate the strong denoising capability of the proposed MD-KSVD. For the measured PD signals, the proposed method yields noise reduction ratio improvements of 24.28% over WT and 87.86% over conventional K-SVD. These results indicate that MD-KSVD provides an effective and robust denoising approach for enhancing PD signals under harsh noise conditions.