Fault Location Method For Overhead Line-cable Hybrid Line Based On LSTM

The urban power supply is widely used overhead line-cable hybrid transmission line.Such transmission lines usually span a large distance,impedance composition is relatively complex,and there is a certain degree of difficulty to find fault points.After the fault has occurred,accurate fault ranging is beneficial to the rapid discovery of fault points and troubleshooting of the line,reducing the economic losses caused by power outages,which is of great significance to ensure the safe operation of transmission lines.Long short term memory is one of the most advanced networks for processing time series,which is widely used in fault diagnosis,image recognition,load prediction,and many other fields.In the field of transmission line fault ranging,scholars at home and abroad have done a lot of research on LSTM networks and achieved some valuable research results,but there are generally problems of complex implementation and poor fault tolerance of LSTM network ranging algorithms.To address these problems,this paper carries out research work related to LSTM ranging algorithms to improve the accuracy and reliability of hybrid line fault location.First,the Dropout layer and pruning mechanism are introduced to optimize the conventional LSTM network structure,which solves the problem of overfitting in the network training process.The Adam adaptive algorithm is used to train the network to speed up parameter optimization.The root mean square error fluctuation range of the optimized LSTM network is small,the loss value decreases rapidly,and the effect of training and learning is better than that of the conventional LSTM network.Secondly,according to the structural characteristics of the hybrid line,a fault interval discrimination method based on the wavelet energy ratio is proposed.The interval judgment is made according to the wavelet energy difference of the bus voltage at both ends when the fault occurs in different sections,and the fault can be preliminarily judged while judging the fault interval distance,PSCAD/MATLAB simulation results prove the feasibility of the method.A hybrid line fault-ranging method based on LSTM is proposed based on accurate fault interval discrimination.The line mode voltage signals on the bus side of the line are collected under different fault types;the discrete wavelet transform is used to decompose the line mode voltage signals to extract the fault characteristics,and the data are pre-processed to obtain the sample set;the LSTM network is constructed using MATLAB software,and the input samples,output samples The LSTM distance measurement model is obtained by adaptive learning.Finally,to test the comprehensive ranging performance of the network,various types of hybrid line fault samples with different fault points and different transition resistances were selected for ranging simulation verification.The results show that the ranging accuracy of the proposed algorithm is higher than that of the wavelet neural network and support vector machine methods and the method is fault-tolerant and unaffected by the transition resistance,fault type,and initial phase angle of the fault,meeting the requirements of engineering practice for positioning accuracy within 200 m.