Noise Reduction of Partial Discharge Detection of Gas Insulation Switchgear via Time-Frequency Matrix-based Singular Value Decomposition

Partial discharge (PD) has been proven to be an effective way to assess the insulation condition of gas insulation switchgear. However, achieving acceptable levels of sensitivity and accuracy of on-site PD measurement is yet a challenging task, primarily due to various field noise interferences, mainly including white noise and discrete spectrum noise. This paper proposes a novel PD de-noising algorithm based on short-time Fourier transform and singular value decomposition. The time-frequency transformation helps distinguish the functional patterns of the PD signal from any overlapping white noise and discrete spectrum noise. Additionally, the matrix decomposition feature helps segregate and filter out the various signal components based on their significance while eliminating the presence of noise. The de-nosing results from the measured PD signals show the algorithm’s excellent performance than the classical approaches.