Information Fusion Fault Line Selection Method Based On The GA-LVQ Network

The problem of power distribution single-phase ground fault line selection has not been solved. Due to the complexity structure of power distribution network, and all kinds of single line selection methods has its limitations, resulting in the overall election line not very satisfactory. So it is of great significance to research the line selection system of selecting and removing fault line quickly and accurately.The thesis first analyzes the fault characteristics of voltage and current signals when the distribution networks take place single-phase ground fault, and then further puts forward an new optimized neural network information fusion line selection method. Because of the LVQ neural network was simple structure, and was easy to implement and had a better clustering effect, and is widely applied in fault diagnosis field. So first to fault transient characteristics of wavelet modulus maxima line selection method, the paper proposes a algorithm based on genetic algorithm optimization(GA) of LVQ neural network to achieve distribution network ground fault line selection method. Through the simulation verify the GA-LVQ neural network using genetic algorithm to overcome the problem that the LVQ network algorithm is sensitive to the initial weight, which was higher precision of line selection and faster convergence speed. Moreover, due to the use of a single fault characteristics of line selection method has certainly limitation in the practical application, so the information fusion line selection method based on GA-LVQ is proposed through using the fault transient characteristic information, construct the zero sequence transient current traveling wave component method, the fundamental wave method and the fault measurement function of active component method. First by using MATLAB software to establish a distribution network of single-phase ground fault simulation model, the different fault types were simulated. Then zero sequence current signals were collected. Fault features were extracted by successively using three kinds of methods and then line fault situations were analyzed. According to the functions of fault measure, to calculate the fault measure as the samples of GA-LVQ neural network information fusion line selection method. Finally the accuracy of this method of line selection was compared with the accuracy of three kinds of single line selection methods. Simulation results show that the proposed GA-LVQ neural network information fusion for line selection method has higher precision and more obviously advantage.This paper contains the GAto optimize LVQ algorithm and intelligent information fusion algorithm, and presents a complete new method of power distribution ground fault line selection. It is progressive to realize clustering algorithm of the LVQ neural network. By MATLAB language to program and optimize the initial weights of LVQ neural network through genetic algorithm, GA-LVQ network combined with information fusion algorithm is used for fault line selection, which was certainly advancement and broadly application prospect in the single-phase ground fault line selection in distribution network.