Yongpeng Xu

Energy Estimation of Partial Discharge Pulse Signals Based on Noise Parameters

Partial discharge (PD) detection has been proved as an effective tool for insulation condition monitoring of power equipment. The energy of PD pulses is valuable for studying the characteristics of PD activities. This paper proposes a method for estimating the energy of PD pulses in the presence of white noise and narrowband noise. First, a maximum likelihood (ML) estimator of the pulse energy is derived from the probability distribution of the energy spectral coefficients. To implement the ML method, the sampled data are divided into signal frames and noise frames. The noise frames are then utilized for extracting noise parameters using the 3F-C method. Eventually, these noise parameters are applied to the signal frames to find the ML estimate of the pulse energy. To verify the effectiveness of the proposed method, both simulated data and measured data have been processed using the proposed method and the conventional wavelet packet (WP) denoising method. Compared with the WP denoising method, the proposed method has a higher accuracy and is less susceptible to the lengths of the sampling time windows. The advantage of the method is more significant in unfavorable conditions where the signal-to-noise ratio is low and the accurate lengths of the PD pulses are difficult to determine.

Modal analysis of three-phase cable systems based on a modified high-frequency model

A modified three-phase cable model, which takes into account the high-frequency characteristics of the cables and the earth, is proposed for high-frequency signal analysis, e.g., the detection and location of partial discharges in cables. Closed-form expressions of the modal propagation constants are derived based on the modal decomposition theory. A global sensitivity analysis of the modal velocities and attenuations is carried out using Sobols method, which makes it possible to establish the effect of external factors on modal propagation. The results show that the wave propagation characteristics at high frequencies are distinctly different from the case of low frequencies.

Partial discharge pattern recognition of XLPE cables at DC voltage based on the compressed sensing theory

Partial discharge (PD) detection has been widely applied to high voltage cable systems for several decades. In this paper, three kinds of insulation defects in XLPE cables are designed and tested at step-wise DC voltage. The PD developing progress of each defect cable is divided into two stages based on the severity degree of PDs. Based on the compressed sensing (CS) theory, a novel method used for recognizing PD patterns at DC voltage is proposed. Firstly, both the statistical features of PD repetition rate and the norm characteristics of time domain features are extracted to create a high-dimensional feature space. Then each test sample from the feature space is sparsely represented as linear combinations of training samples, and the sufficiently sparse one is obtained via 1-norm minimization. Finally, the PD pattern can be recognized by minimizing the residuals between the test sample and the recovered one. The experimental data is analyzed by the proposed method, and the results show that the patterns of both PD source and PD stage are recognized precisely, when the combination solution of features and the 1-norm minimization algorithm are determined appropriately.

Partial discharge feature extraction through contourlet transform for XLPE cable defect models classification

Partial discharge feature extraction is an important processing procedure in XLPE cable defect models classification. For feature extraction in the wavelet domain, the energies of subbands are usually extracted. However, the energy of one subband is just a specific feature. In this paper, we propose an efficient feature extraction method for XLPE cable defect models classification. In particular, feature vectors are obtained by fuzzy c-means clustering on the contourlet domain as well as using two conventionally extracted features that represent the dispersion degree of contourlet subband coefficients. By investigating these feature vectors, we employ the QGA-BP neural network classifier to perform models classification, and experimental result show that our proposed approach outperforms four current PD classification methods.

The application of multi-sensor technique in switchgear partial discharge precise location

This paper describes the basic theory and detection methods of UHF, ultrasonic and high-frequency on-site detection. With the example of failed switchgear on-site detection in 35kV substation, this paper analyzes the application of multi-sensor partial discharge detection technique in the area of switchgear insulation defects detection. Combined with the on-site analysis and disassembled maintenance, the paper verifies the switchgear detection and locations feasibility and accuracy of multi-sensors technique.