Research On HVDC Fault Location Based On VMD And GA-BP Network Fusion

Due to the obvious advantages of UHVDC technology in power transmission,UHVDC transmission is usually chosen when carrying out long-distance and large-capacity transmission.However,China's geographical complexity makes UHVDC transmission lines vulnerable to faults in the process of power transmission.When line accidents occur,it is possible to locate faults quickly and accurately,which will be used in China's power system.The development of UHVDC transmission line fault location is of great significance,so there is no doubt that it has become the primary object of current research.Firstly,signal processing method,namely variational mode decomposition technology,is adopted to process the line-mode voltage of fault transient information,which solves the disadvantage of the original method in calculating the natural frequency affected by various parameters.Secondly,according to the non-linear relationship between the energy spectrum value of variational mode decomposition and the fault distance,a BP neural network is proposed to deal with the non-linear relationship between them.Infinite approximation,because the traditional BP neural network has the disadvantage of transitional dependence on its own weights and thresholds,this paper adopts the combination of genetic algorithm and BP neural network.The global convergence of genetic algorithm can just make up for the disadvantage of BP neural network dependence on weights and thresholds.Therefore,the application of GA-BP neural network in UHVDC fault location can be improved.Accuracy of fault location.In this paper,the(±800kv)Yun-Guangzhou UHVDC transmission system is taken as the research object,and the model of UHVDC transmission is built by PSCAD software.In this model,different transition resistors and different fault distances are set to simulate and extract the fault voltage,and the fault voltage information is passed through the model.The phase-mode decoupling technique extracts the line-mode voltage,and then extracts the energy value through the variational mode three-mode decomposition.It is used as the input training sample of GA-BP neural network for training and learning.The trained model is used for fault location of UHVDC transmission lines.Through the simulation and verification of UHVDC fault location model,the method based on the integration of VMD and GA-BP neural network can achieve high-precision fault location of transmission lines.The maximum error of fault location can be controlled within 1.1km,and the relative error of fault location can be controlled within 0.1%.