Noise suppression method of transformer local discharge signal based on SVD and EEMD joint correlation coefficient

Partial discharge is a primary means of assessing the insulation condition of power transformers. In response to the mixed noise encountered during on-site detection of power transformers, this paper proposes an improved denoising algorithm using Singular Value Decomposition (SVD) and Ensemble Empirical Mode Decomposition (EEMD) combined with the correlation coefficient method. Firstly, signals containing mixed noise are suppressed for narrowband interference using SVD. The noisy signal is decomposed using SVD, and an improved threshold algorithm is employed to separate the singular values corresponding to narrowband interference, reconstructing them as narrowband interference, and removing them from the contaminated signal, accomplishing the elimination of narrowband interference in mixed noise. Since the singular values corresponding to white noise change slowly, SVD is not suitable for suppression. This paper addresses white noise suppression by first employing EEMD decomposition, selecting the decomposed Intrinsic Mode Function (IMF) components using the correlation coefficient method, applying wavelet transform denoising to high-frequency components, and finally reconstructing the denoised components with low-frequency components, completing the removal of white noise from mixed noise. Finally, through comparison with other existing algorithms, the results indicate that this method exhibits better suppression effects on white noise and narrowband interference, offering improved restoration of partial discharge signals.