Research On Cable Fault On-Line Diagnosis Method Based On Deep Learning Theory And Phase Velocity

With the widespread use of power cables,off-line detection method of cable fault has brought tremendous pressure to the power sector,in order to ensure the safe operation of power grid and reduce the artificial maintenance costs,cable fault diagnosis method should be changed from offline detection to online.At present,the diagnosis of cable fault is still dominated by off-line method,online diagnostic methods are in the stage of exploratory research and many theories exist many problems,in the previous study,the method of traveling wave fault location often mistake the cable voltage or current signal as a traveling wave signal,and the object of study is a single cable.It not meet the requirements of cable fault diagnosis technology in the new situation.In view of the above problems,this paper establishes the simulation model of underground cable distribution system for collecting the voltage and current signals of different fault types in different situations.The concept of depth learning is introduced to analyze the cable fault,and the fault distance is calculated by the phase difference of double ended signal waveform.The following research results are obtained in this paper:(1)A simulation model of three-phase power supply system with 16 underground power cables is established,different types of faults are set up in different position cables,and the various faults of cable are simulated from the theoretical point of view.The model can make up the shortcomings of poor adjustability and data in the actual system.The running status of the cable and the rule of fault occurrence are reflected by using a large amount of data.So this model can be used as an effective complement to the fault detection technology in the actual operation of the cable.(2)The depth belief network(DBN)and convolution neural network(CNN)based on depth learning theory are set up for cable fault identification.The depth neural network can complete the classification of fault signal automatically and locate the fault accurately on the specific cable by using a large number of fault data.(3)The mature method of traveling wave is abandoned and the phase velocity method is used to obtain fault distance.The mathematical expression between phase difference and the fault distance is deduced according to the phase difference of the actual voltage and current waveform of the cable,the required data acquisition and calculation process is simple.(4)A visual detection-system is designed based on MATLAB-GUI,the simulation model,cable fault identification,fault distance calculation and waveform display are integrated into the system,the fault setting is simple and easy to identify.Finally,the system was tested and applied.The DBN&CNN based on depth learning and traditional shallow neural network is compared by experiments,the average recognition accuracy of DBN and CNN is 89%and 93%,the traditional BP,RBF and SVM is 50.8%,67%and 83%.Compared with traditional diagnostic methods,the method proposed in this paper reflect the cable running status and fault rule by using the big data,the fault recognition accuracy and fault location accuracy are improved significantly,it is an effective supplement to the fault diagnosis technology in the cable actual operation,it has a certain theoretical significance and use value.