For removing effectively the white noise interference in the partial discharge ultrasonic signals of transformer and improving the accuracy of subsequent partial discharge pattern recognition and location

a kind of denoising method of partial discharge ultrasonic signal of transformer based on improved wavelet threshold and complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is proposed. Firstly

a series of intrinsic mode functions (

IMF

) from high frequency to low frequency are obtained by CEEMDAN decomposition of discharge signal. Then

a multi-scale permutation entropy (MPE) algorithm is used to calculate the permutation entropy (

PE

) of each

IMF

component

and the denoising threshold and effective threshold of the signal are determined in accordance with the permutation entropy of each

IMF

.

IMF

components above the denoising threshold are denoised by the improved wavelet threshold method and

IMF

components below the effective threshold are regarded as baseline drift. Finally

the de-noised ultrasonic signal is obtained by reconstructing the de-noised component and the residual component. Both simulation

and experimental results show that the denoising algorithm proposed in this paper improves greatly the signal-to-noise ratio of the signal

and retains the effective information of the original ultrasonic signal

which is of great significance to improve the accuracy of partial discharge pattern recognition and location using ultrasonic signals.