Research On Method Of Fault Recognition And Location For Distribution Line

With the prompt developing of China’s economy,the power system has occupied an increasingly critical status in social growth.In the construction of the power system,the low-voltage distribution network has also developed rapidly.As the most faulty part of the power system,distribution lines seriously interfere with the stable working state of the power system.Therefore,it is of significant actual importance to recognize the variety of fault and decide the position of the fault in a appropriate and correct method.This article first analyzes the theory of the overall average empirical mode decomposition(EEMD)transformation and the singular value decomposition algorithm.Use EEMD to decompose the fault signal of the line and obtain the wavelet coefficients corresponding to each scale to construct the signal matrix.Then,the singular value decomposition(SVD)is used to process the signal matrix to obtain the singular value eigenvalue representing the complexity of the signal.Then,in view of the misjudgment of the intelligent algorithm when processing the distribution line unbalanced data set,an improved restricted radius oversampling algorithm(LR-SMOTE)is proposed.This algorithm solves the problems of over-fitting and data marginalization in the traditional sampling algorithm(SMOTE)by processing unbalanced sample data sets.furthermore,because the fault signal is frequently influenced by noise obstruction,it is hard to judge the particular fault line.A denoising means combining variational modal decomposition(VMD)and singular value decomposition(SVD)is advised.which can maintain the characteristics of the original signal waveform.On the premise of eliminating the influence of noise signal on the original fault signal waveform,a fault line extract way based on the relative entropy of VMD energy is recommended.The measured location results show that the fault line selection means advised in this paper has a high carefulness frequency and is not influenced by different fault element s.Finally,considering that the topological structure of the distribution network is relatively complex and the working status is often changed,the traditional use of a single feature quantity for fault location will have a poor effect.Therefore,this paper decided to combine multiple characteristic values to determine the location of the fault on the faulty line.In this paper,the localization algorithm processes the fault zerosequence current signal through the complete set empirical mode decomposition(CEEMDAN)method,and then combines the relative energy entropy and the singular value Euclidean distance to construct a comprehensive fault measurement function reflecting the degree of fault occurrence.This method reduces the blind zone of segment positioning and improves the accuracy of segment positioning.