A Novel Adaptive EEMD Method for Switchgear Partial Discharge Signal Denoising

Abstract

The elimination of a variety of noises such as the narrow-band interference in the detection of partial discharge (PD) signals in switchgear is an intractable issue. Furthermore, the self-adaptation in the denoising process is weak. A partial discharge-based novel adaptive ensemble empirical mode decomposition (Novel Adaptive EEMD, NAEEMD) method is proposed in this paper for noise reduction. First, the signal is decomposed using the EEMD, only the first-order natural mode is decomposed until the signal margin reaches the EEMD decomposed termination condition. After removing the first-order mode, noise is added to the residual signal, and the remaining signal components are decomposed in the next stage. At last, the intrinsic mode function (IMF) of the noise reduction reconstruction is adaptively selected. The latter is accomplished by combining the energy density and the average period of the IMF correlation coefficient method. Meanwhile, the proposed method provides a new strategy for pre-processing the PD signal of the switchgear. The outcomes of the proposed NAEEMD de-noising method have been compared with the conventional wavelet denoising algorithm (WDA) and EMD-based threshold denoising for validation. The simulation results showed a good denoising effect and effectiveness of the proposed method compared to the WDA and EMD-based threshold denoising. Furthermore, an experimental simulation utilizing actual switchgear PD signal has been performed to verify the noise reduction effectiveness of the proposed method.INDEX TERMS Switchgear, partial discharge, NAEEMD, narrow-band interference, denoising, wavelet transform.

Denoising of Radio Frequency Partial Discharge Signals Using Artificial Neural Network

“…The structure of the employed ANN includes a feed-forward network, in the input and output layers, comprised of sigmoid and linear activation functions, respectively. As shown in Figure 4, a series of samples, which are the ANN input vector, is approximated using an appropriate curve and the output of ANN corresponds to the least square error and obtained by Equation (9). Thus, the acquired curve in the process is considered as the denoised signal.…”

“…Energies 2019, 12, 3485 6 of 14 appropriate curve and the output of ANN corresponds to the least square error and obtained by Equation (9). Thus, the acquired curve in the process is considered as the denoised signal.…”

“…Wideband noises, sometimes called background noises, have a stochastic nature and depend on the measuring system as more sensitive measuring systems are more prone to wideband noise [3,5]. To remove such noise, digital signal processing algorithms-including mathematical morphology [6,7], empirical mode decomposition [8,9], and wavelet transform [2,10]-can be exploited. Mathematical morphology is a time-domain and effective algorithm with a low computational burden but the determination of the type and length of the structure element has always been a challenge [6].…”

“…Empirical mode decomposition is a time-domain algorithm that is recently proposed for white noise suppression. Although this method is able to find the level of signal decomposition through a self-adaptive approach, the computational burden to solve the problem of mode mixing, as well as thresholding process, are the obstacles that are yet not completely solved [9]. Being one of the most prevalent algorithms, wavelet transform (WT) has long been exploited in PD signal denoising.…”

An Improved Denoising Method for Partial Discharge Signals Contaminated by White Noise Based on Adaptive Short-Time Singular Value Decomposition

“…Chan et al [11] proposed a hybrid method called ensemble EMD (EEMD) to overcome mode mixing and avoid end effect in EMD. Moreover, Jin et al [12] used a novel method called novel adaptive EEMD (NAEEMD) to decrease the complexity and computational time of EEMD when white noise was superimposed multiple times in the original signal. To prevent these deficiencies in EMD, methods based on SVD have been proposed [13,14].…”