The partial discharge (PD) classification for electric power equipment based on machine learning algorithms often leads to insufficient generalization ability and low recognition accuracy. To solve the problem, this paper develops an improved Wasserstein generative adversarial network with gradient penalty (WGAN-GP) based data augmentation model. The improved WGAN-GP model can generate data samples to supplement the low-data input set in PD source classification. Firstly, an improved WGAN-GP model with conditional generation is trained and various new data samples are generated. Then the new data samples are utilized to expand the raw data set. Finally, the expanded data set is trained to get a new PD classifier. Experimental results demonstrate the proposed model can generate new high-quality data samples more stably. Moreover, the proposed method can suppress the over-fitting risk caused by low-data or imbalanced data distributions and the classification accuracy is effectively improved.

This article proposes a cable defect location method based on the power spectral density (PSD) to improve the accuracy of the existing frequency-domain reflection (FDR) method. The principle of the proposed method is that empirical mode decomposition (EMD) is used to decompose the envelope of the reflection coefficient into the intrinsic mode function (IMF) of each insulation defect point and the real part of each IMF signal is windowed. Then, according to the characteristics of a random signal with a definite auto-correlation function, PSD analysis is performed on the IMF signal with limited power. In theory, the power spectrum of a stationary random process is a deterministic function, which transforms the analysis of uncertain signals into the analysis of deterministic functions from the mathematical principle. And combined with the FDR method, localizes the cable insulation defects. Finally, the feasibility and effectiveness of the proposed method are verified by simulation and experiment.

The urgent need for power transmission is the reason for leading research on the safer operation of high‐voltage cables. These high‐voltage cables are emerging as an efficient technology for underground power transmission. However, the electrochemical corrosion of aluminium (Al) metal sheaths in these cables is a common and challenging degradation process. Herein, the corrosion mechanisms and electrochemical analysis are investigated by using scanning electron microscopy (SEM), Fourier transform infrared spectroscopy (FTIR), electrochemical impedance spectroscopy (EIS) and laboratory‐designed electrochemical characterisation techniques. In the corrosion experiments, the authors evaluated the corrosion rate by measuring the release rate of hydrogen gas and explored the different roles of sodium polyacrylate (NaPA) during the corrosion process. It is found that the complexation reactions between NaPA and Al inhibited corrosion while increased the resistance of the buffer layer. The proposed mechanisms of corrosion in this study can improve the lifespan and sustainability of high‐voltage power transmission.

Due to the spurious peaks caused by multi-reflection and the randomness of the cable transfer function, the frequency domain reflection (FDR) method based on Fourier transform (FT) is inaccurate, even resulting in misjudging the soft faults of the cable. Thus, in this paper, a frequency sweep location method based on multiple signal classification (MUSIC) pseudospectrum is proposed to accurately diagnose and locate the soft faults. The idea of the method is to extend the cable transfer function into an array signal and estimate the MUSIC-pseudospectrum of the transfer function, which can transform the analysis of uncertain signals into the analysis of deterministic functions. Thus, the impact of the spurious peaks and the randomness of the cable transfer function can be eliminated. First, the transfer function of the cable is obtained by frequency sweep. Then the soft fault information in the transfer function is extracted and reconstructed by variational mode decomposition (VMD), which has a good ability to divide the harmonic frequency band. The MUSIC-pseudospectrum of the reconstructed signal is estimated to locate the cable soft faults. At last, the feasibility and effectiveness of the proposed method were verified by testing on a laboratory cable and field underground cable.

The partial discharge (PD) in inverter-fed motors generated from high frequency and short rise time impulsive voltage are more complex than traditional sinusoidal voltage operated motors due to the noise initiated from fast switching power electronics devices. Available PD sensors and related technology for DC and sinusoidal voltage cannot be used at repetitive impulsive voltage conditions. This work aims to report the design of antenna with optimized geometry and satisfied parameters to detect PD for impulsive voltage. The antenna is verified and the influence of both impulse voltage parameters and test configuration on PD features are discussed. The envelope detection technique after the filter in the frequency domain was proposed to reduce the complexity and cost of PD test hardware under impulsive voltages with fast rise times. The frequency domain energy of PD from motor insulation is mainly distributed in the range of 0.6-1.8 GHz. By increasing the amplitude and frequency of input impulsive voltage, the range of PD energy spectrum is enlarged to (1.5-1.8 GHz) and by decreasing input amplitude the PD signal strength decays abruptly. Moreover, the high-frequency distribution of PD energy attenuates more severely with the distance and the electromagnetic energy of PD decays nonlinearly with the increase of propagation distance. Therefore, under the premise of satisfying the safety test, selecting the test distance below 15 cm can ensure that the test results have a high signal-to-noise ratio and signal integrity. The experience reported in this study could provide suggestions for off-line PD measurements for inverter-fed motor insulation evaluations for manufacturers.

The polarization and depolarization current (PDC) measurement has been widely used to assess the aging degree of cross-linked polyethylene (XLPE) cables. However, the aging type identification of XLPE cables has not been fully considered. In this study we conduct PDC measurement on thermal-aged and water tree-aged XLPE cable samples respectively, upon which a graphical approach to assess the insulation condition and distinguish the aging type of XLPE cables is proposed. First, to designate the evaluation indexes of the insulation condition, multiple PDC measurement based dielectric parameters are considered. Then, to reflect the sensitivity of each evaluation index to the condition of XLPE cable, a combined weighting method based on the group fuzzy analytic hierarchy process (GFAHP) and entropy method (EM) is introduced. Furthermore, the improved radar chart method (RCM) based graphical approach is proposed to comprehensively and intuitively characterize the evaluation indexes. Finally, the aging degree of the XLPE cable samples is quantitatively assessed by calculating the area of the polygon in the chart. Simultaneously, utilizing the position of the areal coordinate and the shape characteristic parameter of the polygon, the thermal-aged and water tree-aged samples are effectively distinguished.