Research On High Voltage Cable Buffer Layer Ablation Fault Identification Based On Frequency Domain Impedance Spectrum And Machine Learning

Cross linked polyethylene(XLPE)insulated cables have been widely used worldwide due to their advantages such as lightweight,easy installation,excellent electrical and mechanical properties,and heat resistance.The erosion fault of the buffer layer in high-voltage XLPE insulated cables is a new type of fault that has emerged both domestically and internationally in recent years,with different fault locations and characteristics from traditional cable faults.This fault will form continuous erosion marks at the fault point,and if not detected and dealt with in a timely manner,it may puncture the cable line and seriously affect the safe and stable operation of the power system.At present,the research on the mechanism and detection methods of this fault both domestically and internationally is not mature enough.Therefore,studying the fault mechanism and detection and identification methods of buffer layer erosion fault can timely detect and handle cable faults caused by such defects,which has important engineering application value.On the basis of existing research on buffer layer erosion faults both domestically and internationally,this article conducts relevant research on the mechanism,fault detection and identification of buffer layer erosion faults.The main content is as follows:The fault mechanism of buffer layer ablation failure was studied and analyzed.Selecting 110kV high-voltage XLPE coaxial cables as the research object,combined with the actual structure,size,and material parameters of high-voltage cables,two-dimensional and threedimensional models of high-voltage cables with water blocking buffer layer structures were established.The finite element method was used to analyze the built model in the physical field of electromagnetic field temperature field coupling.By setting whether the buffer layer is affected by moisture and whether the buffer layer has good contact with the wrinkled aluminum sheath,the erosion fault of the buffer layer is simulated.The temperature rise and electric field distribution caused by the core current and capacitive current are observed.The results show that the erosion fault of the buffer layer is caused by the joint effect of moisture on the buffer layer and the poor contact degree between the buffer layer and the wrinkled aluminum sheath.To more intuitively represent the variation of temperature rise with the length of the contact defect,a functional relationship between the distance of the contact defect and the temperature rise of the buffer layer under dry and damp conditions was fitted using simulation data.In order to overcome the limitations of existing methods for detecting buffer layer ablation faults,the frequency domain impedance spectrum method of the cable head is used to detect buffer layer faults.Based on the theory of transmission lines,a distributed parameter model of cables with normal and local defects was constructed.Different types of local defects,including buffer layer ablation faults,were simulated using changes in insulation material parameters and fault factors.The frequency domain impedance spectra of cable heads with different defect types were simulated in the 0-100MHz frequency band,and different lengths,types The characteristics of the impedance spectrum of the cable head corresponding to local defects of different severity were analyzed,demonstrating the feasibility of using impedance spectrum method for detecting cable buffer layer ablation faults.Aiming at the characteristic that the frequency domain impedance spectrum of cable head is sensitive to capacitive defects but difficult to distinguish the types of defects,a cable buffer ablation fault identification method based on the frequency domain impedance spectrum of cable head and Multilayer perceptron neural network is proposed.Simulate different types of local defects by setting different fault locations and lengths,and extract corresponding amplitude spectrum data to establish a dataset;Six different machine learning classification and recognition algorithms are selected to train and test the data sets before and after preprocessing,analyze the necessity and effectiveness of the data preprocessing method used in this paper,and compare from multiple perspectives.Finally,Multilayer perceptron MLP neural network is used as the fault classification and recognition network to achieve effective identification of buffer layer ablation fault,The effectiveness and superiority of the proposed method for identifying erosion faults in high-voltage cable buffer layers have been verified,providing a new approach for identifying erosion faults in cable buffer layers.