On Online Measurement Of The Power Cable's Partial Discharge

In recent years,China's demand for electricity has increased significantly,so the power cables have been widely used in power transmission systems at all levels.During the longterm operation of the cable,there may be some safety hazards.Studies have shown that the partial discharge phenomenon of the cable can reflect the degree of deterioration of the cable insulation material to some extent,and can show the defects of the cable.At present,the oscillating wave technology is widely used to detect the off-line local discharge of power cables.However,China's large power grid system makes it very difficult to carry out test under the loss of electricity supply,so this technology is restricted to a certain extent.At the same time,offline detection cannot simulate the actual running state of the cable.Therefore,applying online measurement of the power cable's partial discharge can detect the cable abnormality in advance,can effectively prevent the cable failure and increase the service time of the cable,and has a great research value.In this thesis,the on-line detection technology of partial discharge of power cable is deeply studied.Firstly,the propagation law of partial discharge pulse signal in power cable is analyzed.Then,the front-end sensor and data acquisition system are designed.Finally,aimed at the phenomenon that the coexistence of pulsed interference signal and PD signal,a cluster recognition algorithm for partial discharge signals is proposed.Firstly,this thesis studies the research background and research status of partial discharge online measurement technology,expounds the source of noise and cluster recognition technology in partial discharge detection,analyzes the causes of partial discharge phenomenon and discharge mechanism,and gives five mathematical models to simulate it according to the characteristics of PD signal.In order to study the transmission characteristics of partial discharge signal in power cable,this thesis uses double exponential decay model as PD pulse signal and establishs distributed parameter line model of power cable in SIMULINK.According to this model,the influence of cable length and distribution parameters on signal transmission are analyzed in this thesis.Secondly,after a comprehensive comparison of various PD detection methods,this thesis uses electromagnetic coupling method to realize the on-line detection of PD signals.This thesis designs a self-integral high-frequency current sensor based on Rogowski coil and builds a signal acquisition system.According to the principle of different partial discharge phenomena,three typical models of partial discharge were designed and made,and the partial discharge test platform was set up in the laboratory.For the current sensor,the parameters of the winding and the integral resistance of the coil are determined by analyzing the parameters in MATLAB,and the performance of the sensor is tested at different frequencies.For the signal acquisition system,the hardware part adopts the design scheme of high sampling rate and high precision oscilloscope combined with portable computer.The software part uses LABVIEW and MATLAB mixed programming to realize data acquisition and processing.LABVIEW is used to realize the communication between oscilloscope and computer,the display of the waveform,data export and so on.MATLAB is used to implement signal processing.Finally,for the coexistence of pulsed interference signals and PD signals,this thesis proposes a cluster recognition algorithm for partial discharge signals,which mainly includes pulse extraction,feature parameter extraction,cluster analysis and effectiveness evaluation.This algorithm adaptively determines the threshold by the maximum inter-class variance method firstly,and extracts the pulse signal by the combination of amplitude-time double thresholds method and the time domain energy method.Then the processing of the acquired pulse signal to obtain the feature vector is carried out according to the synchronous multi-channel method.Next,the fuzzy C-means clustering algorithm is used to achieve clustering separation of useful signal and noise signal.The clustering effect can be analyzed and tested by Phase Resolved Partial Discharge(PRPD) distribution and two-parameter Weibull distribution function fitting.The experimental results show that compared with the equivalent time-frequency method,the feature parameters obtained by the proposed algorithm are more capable of representing the signal characteristics;at the same time,the algorithm uses the subtractive clustering to initialize clustering center and can effectively cluster various discharge signals,so as to realize single class discharge signal type recognition.