Research On Fault Diagnosis Technology Of Distribution Network With Distributed Generation

As the “last mile” of power supply for users in the power system,the distribution network undertakes the important task of power distribution,measurement and protection,and is the most important basis for ensuring people’s livelihood.The distribution network has a wide distribution of lines and a high probability of faults.Once a fault occurs,it can cause mild power outages and severe impacts on national livelihoods.Therefore,accurately identifying the type of fault,determining the fault line and fault section can greatly shorten the fault processing time,quickly restore power supply,and will have extremely important significance in reducing economic losses and improving reliability.Based on sparse representation theory,this thesis studies the problem of distribution network fault type recognition,fault line selection and fault section location.The main contents are as follows:Starting from the basic concepts of compressed sensing theory,this thesis analyzes and studies the basic concepts and mathematical models of sparse representation theory,introduces two core problems in sparse representation theory: sparse coefficient solution and over complete dictionary construction,and introduces in detail the common algorithms for these two types of problems.Through comparative analysis,Orthogonal Matching Pursuit(OMP)algorithm was selected as the sparse coefficient solution in this thesis,and K-Singular Value Decomposition(K-SVD)algorithm was selected as the dictionary construction algorithm.It lays a theoretical foundation for the following fault diagnosis methods.In view of the difficulty in fault feature screening and complexity of threshold setting in distribution network fault type recognition methods,a distribution network fault type recognition method based on Sparse Representation based Classification(SRC)was studied.Using the time-domain characteristics of the three-phase voltage and zero sequence voltage after a fault as the basis for fault classification,the K-SVD algorithm is used to learn the characteristic information of various fault signals,and an over-complete dictionary matching the essential characteristics of various faults was constructed;use the OMP algorithm to sparsely decompose the voltage signal,and combine the SRC algorithm to compare the residual between the reconstructed signal and the test signal.The test signal is classified as the category corresponding to the minimum residual,Furthermore,the identification of fault types in the distribution network can be achieved.In addition,on the basis of traditional distribution network fault type identification models,Distributed Generation(DG)is connected,and the applicability of this method in distribution networks with DG is verified using MATLAB/Simulink simulation software.In order to solve the problems of small single-phase grounding fault current and difficult to extract criterion feature quantity in low current grounding system,an adaptive dictionary based sparse representation method for small current grounding fault line selection is studied.Firstly,the sparse decomposition of the adaptive dictionary is applied to the three-phase voltage,zero sequence voltage,and zero sequence current of each feeder line on the busbar after a small current grounding fault.Then,the maximum sparse coefficient of the zero sequence voltage is used to determine whether a single-phase grounding fault has occurred.The maximum sparse coefficient of the faulty three-phase voltage is used to determine the faulty phase.The maximum sparse coefficient of the zero sequence current of the feeder line is used to construct a small current grounding fault line selection criterion,achieving fault line selection.Finally,MATLAB/Simulink simulation software was used to complete the verification.In view of the problem that most of the current methods of distribution network fault line selection and section location are only suitable for single-phase earth short circuit fault a sparse representation based fault line selection and section location method is studied.This method utilizes the three-phase current and zero sequence current of each feeder of the bus before and after the fault,as well as the measurement points of the fault line,as input electrical quantities.The OMP algorithm solves for its sparse representation coefficient,and utilizes the maximum sparse coefficient of the fault line current splicing signal to always be greater than that of the non-fault line selected fault line;Take out the faulty line,use each measuring point on the line as a node,and the area enclosed by adjacent measuring points as a positioning interval.Construct a network description matrix of the faulty line based on the matrix algorithm.By comparing the maximum sparse coefficient variation law of the current splicing signal at each measuring point,select the sparse coefficient variation function with the minimum amplitude change when single phase grounding,and based on this,establish a fault information matrix,Multiply the network description matrix with the fault information matrix and normalize it to obtain the fault discrimination matrix,which then identifies the fault section.Finally,on the basis of traditional distribution network fault line selection and section positioning models,the applicability of this method in distribution networks with DG is verified using MATLAB/Simulink simulation software.By comprehensively utilizing the electrical quantity information and switch quantity information of the distribution network,and utilizing multi-agent(Agent)technology to comprehensively coordinate multi-source data,a fault diagnosis model is constructed.Genetic algorithm is used to transform the fault diagnosis problem into a 0-1 mathematical optimization problem,and a scheme for using multi-source heterogeneous data fusion for fault diagnosis in DG containing distribution networks is designed.