Research In Features Extraction And Fault Selection For Small Current Grounding System Of Single-phase Ground Fault

Currently, small current grounding system is mostly used in medium and low voltage distribution network in our country. When a small current grounding system in the event of failure, how to accurately select the fault line, and timely termination failure relates to the stability of the grid. So, the research of fault line selection of small current grounding system has been becoming a hot topic of current research.First, the status of small current grounding system fault line selection methods was introduced, and the signal characteristics of single-phase ground fault occurs in the small current grounding system, and a number of factors affecting the fault line were analyzed.Considering the disadvantage of end effects of the local mean decomposition theory study, the minimum squared distance-related was given in this paper proposes to improve the algorithm, and end effect was suppressed effectively. An improved local mean decomposition was used to analyze the zero-sequence current signal of small current grounding system and the feature amount of fault line selection was got. After analyzing the theory of support vector machine, least squares support vector machine was used to classify the feature quantity obtained through local mean decomposition, the least squares support vector machines combined with improved local mean decomposition as the first method of small current grounding fault line selection, this method in a larger amount of data that can simultaneously handle multiple sets of data, decresing the time of line selection.The correlation dimension could reflect the feature quantity of the system state, analyze the fault condition quantitatively, and improve the capabilities of fault diagnosis. The improved local mean decomposition combined with correlation dimension, as the second method of small current grounding fault line selection, this method is more suitable for fault line selection is limited only single group or several groups of data.Finally, by using power system simulation software to build a model of small current grounding system, then extract the zero-sequence current signals. Using the two methods proposed for experimental verification, the simulation experiment for small current grounding fault system showed that both methods could select fault line under the condition of small signal amplitude, less data fault information and other unfavorable conditions. Compared least squares support vector machine with radial basis function neural network(RBF), experimental results verify the advantages of LS-SVM.