CEEMDAN and Wavelet Thresholding for HVDC Power Quality Denoising

Accurate perception of high-voltage direct current transmission systems is critical for grid security and stability. In such systems, strong corona-induced impulsive noise often obscures transient disturbance features, while conventional denoising methods lead to waveform distortion and inaccurate modal separation. To address this issue, a hybrid method combining adaptive modal decomposition and improved wavelet thresholding is proposed. A geometric impulsiveness index is introduced to distinguish noise-dominated and signal-dominated components, followed by nonlinear filtering of noisy modes. Results show that the method suppresses oscillations and amplitude distortion while preserving transient features such as commutation notches. Under a 10 dB noise condition, the signal-to-noise ratio is improved by 17.69 dB on average, demonstrating enhanced robustness and reconstruction accuracy in complex electromagnetic environments.