Study On Line Selection Of Low Current Ground Fault Based On Improved BP Neural Network

As the connection hub between the power system and users,the distribution network has a complex operating environment and a wide coverage area,with a high failure rate during operation,and more than 80% of the faults are single-phase grounding faults.If the grounding fault cannot be removed in time,causing the system to run with a fault for a long time,it may further develop the fault,affect the safe operation of the system,and even threaten personal safety.Therefore,it is of great significance to quickly determine the specific fault lines for the safe and stable operation of the power grid and to ensure the personal safety of relevant personnel.This article summarizes the existing fault line selection methods of the distribution network,points out the problems of the current methods,and analyzes the transient and steady-state electrical characteristics of the small current grounding system.In the Matlab /Simulink environment,a small current grounding system model of 10 k V distribution network is built.On this basis,a grounding arc model and a high resistance grounding fault model are built.The influence of different grounding fault types,initial phase angle,fault position and transition resistance resistance value on the fault current is simulated and analyzed.In this paper,the transient frequency band energy,the transient active power component and the transient integral parameter of the zero-sequence current are taken as the characteristic parameters of fault line selection.The VMD algorithm for extracting the energy of the transient frequency band is improved,the sample entropy is selected as the fitness function,and the optimal parameter combination of the VMD is quickly determined by the genetic algorithm optimization.The improved VMD algorithm is used to decompose the noisy zero-sequence current.The results show that the improved algorithm can effectively decompose the signal,avoid the problems of excessive decomposition and modal aliasing in the EMD algorithm,and has the function of filtering and denoising.This paper uses a route selection method using gray wolf algorithm to optimize BP neural network for multi-feature fusion.In order to solve the problems of slow convergence speed and easy to fall into local optimum of BP neural network,gray wolf algorithm is used to optimize BP neural network,and then the characteristic data is input into the trained GWO-BP neural network model to identify the fault lines.Compared with the results of single feature selection and fusion selection before and after optimization,as well as the convergence speed and output error of the algorithm,the results show that the fusion line selection method of GWO-BP neural network can effectively improve the accuracy of line selection and is more efficient.Finally,the proposed method is verified by the measured data,and the results show that the proposed method can accurately identify the fault lines.There are 40 figures,9 tables and 88 references in this paper.