Monitoring And Recognition Of Partial Discharge In Power Cable

With the power cable is more and more widely used in power network, the cable insulation condition has become a problem of concern to everyone, partial discharge is one of the important external manifestations of the deterioration of insulation condition. At the same time, partial discharge will lead to further deterioration of insulation. In order to detect insulation defects of the cable, to ensure the normal operation of the power grid, it is necessary to carry out on-line monitoring of partial discharge of cable. The domestic and foreign scholars have done a lot of research on cable partial discharge on-line monitoring. Based on this, several key links are studied and explored in this thesis.This thesis introduces the cable partial discharge mechanism firstly, secondly, introduces several commonly used detection methods, analysis of the nature of the field test for the existence of interference source and the common interference, according to the characteristics of partial discharge, the four PD mathematical model is selected. Which provide a basis for the simulation signal. Further, we analyzed the reasons of cable insulation defects, and 4 kinds of typical insulation defects were simulated test. In this way we obtain the discharge signals of 4 kinds of insulation defects under a large, which provide a sample for the pattern recognition of partial discharge in power cable. The first key part of the on-line monitoring technology of the cable partial discharge is to detect the signal pretreatment. Which is to suppress the noise in the discharge signal. In this thesis, we mainly focus on the elimination of background noise and periodic narrowband noise. According to the characteristics of cable partial discharge signal and noise signal, we select the method of wavelet packet analysis deal with white noise, Wavelet packet threshold method is commonly used to denoising and how to select the wavelet base and threshold processing is crucial for wavelet packet denoising, In this paper, we start with the threshold function, through the analysis of the traditional threshold function method, using an improved threshold function of wavelet packet denoising algorithm. The simulation test and practical application shows that the improved the method has better denoising effect than traditional algorithm. Considering the frequency of periodic narrowband interference, they are suitable to be analyzed in frequency domain. In this paper, FFT technology is used to suppress narrowband interference. Simulation results show that FFT technology has advantages in the removal of periodic narrowband interference.The second key links in the on-line detection of the partial discharge of the cable are the identification of the partial discharge. The detected cable partial discharge signals to correctly identify the type of the discharge to make quick judgments, take corresponding measures are critical. Pattern recognition of the input object is often not discharge signal itself, and the need to us from the partial discharge signals extracted feature vectors as a pattern recognition input, so it is very important to characterize the characteristics of different partial discharge signals. In this paper, according to the characteristics of the partial discharge signal of the cable, the discharge signal is decomposed by empirical mode decomposition and wavelet packet decomposition. Then calculate intrinsic mode function of the energy coefficient and wavelet packet decomposition coefficient of the frequency band energy as the feature vector of discharge signal. At last, the algorithm of BP neural network and fuzzy c-means clustering algorithm for recognition. The results show that the energy of the intrinsic mode component is more able to characterize the electrical signal characteristics; On the same feature vector, BP neural network algorithm recognition rate is higher than the overall fuzzy C- mean clustering algorithm.