Denoising method for partial discharge signals under high frequency working conditions based on CFOA-VMD and multi-scale wavelet thresholding

To address the issues of overlap between high-frequency power supply background noise and partial discharge signals, the reliance on empirical determination of optimization parameters in existing variational mode decomposition (VMD) algorithms, and the lengthy optimization time of other intelligent algorithms, a high-frequency partial discharge signal denoising method based on Catch Fish Optimization Algorithm-variational mode decomposition (CFOA-VMD) combined with multiscale wavelet thresholding is proposed. First, the characteristics of the high-frequency power supply background noise are analyzed using time-frequency domain methods; next, CFOA is employed to optimize the VMD parameters and perform signal decomposition; then, kurtosis values of each mode component are calculated to distinguish noise from effective signals; subsequently, multiscale wavelet thresholding is applied for secondary denoising of the retained effective components; finally, the denoised partial discharge signal is reconstructed. Simulation and experimental comparisons demonstrate that the proposed method effectively suppresses high-frequency noise in partial discharge signals, improving the signal-to-noise ratio by 23.7%, increasing the noise rejection ratio by 10%, and reducing denoising time by 80%. This research holds significant engineering application value in the field of on-site high-frequency partial discharge noise detection and elimination.