Research On Short Circuit Fault Diagnosis Method Of Distribution Network Based On Multi-source Information Fusion

Accurate and rapid research and judgement of short-circuit faults in distribution networks is of great significance for improving the reliability of power supply and ensuring the safe operation of distribution networks.The traditional fault diagnosis mainly depends on the information of current and voltage or the action of switch.It is difficult to identify complex fault with the fault information is single.Thus,this thesis studies phase to phase fault location method of distribution network relied on the fusion of electrical information and switch action information.Aiming at the problems of numerous nodes in distribution network and relatively few installations of electrical information collection equipment,a fault judgment method of distribution network based on compressed sensing algorithm is proposed.This method uses the electrical information collected by a small number of nodes to obtain a sparse voltage sag matrix,and combines the node impedance matrix to construct an underdetermined equation.The compressed sensing algorithm is used to solve the injection current of all nodes,so as to locate the fault area.Since the number of measurement nodes is much smaller than the total number of nodes,only one reconstruction has a large error,so two reconstructions are carried out: the first reconstruction determines the measurement nodes adjust to the fault area and the approximate fault range,and the second reconstruction determine the fault area.The examples are given to test the accuracy of the method,and the results are less affected by the transition resistance.Aiming at the problems of mal-operation,refusal to operate and data distortion in switching information,it is proposed a neural network model to diagnosis fault of distribution network.The model uses the action of protection and circuit breaker as the input,and the failure probability of each region in the network as the output,and establishes the mapping between input and output to realize fault location.Meanwhile,because the BP network model is easy to be caught in the problem of local minimum,the Firefly algorithm is used to optimize the parameter of BP model and the accuracy is raised.Besides,the simple and complex distribution network models are set up to verify the effectiveness of the method,and the performance of the BP neural network model and the optimized model of the firefly algorithm in terms of fault location accuracy and fault tolerance are compared.Because of the poor anti-interference capability of the single source,this paper uses fusion technology to fuse the results of electrical information and switching information to get the final result.Due to the problems of evidence conflict in the traditional DS evidence theory,this paper improves the probability assignment function and fusion rules of the evidence body by combining the characteristics of electrical faults and switch faults,which is more reasonable Then the final fault location result is obtained according to the decision model.Through example analysis,the proposed method can locate faults accurately.Compared with single information source,it can deal with the situation of inaccurate and incomplete fault information.The short-circuit fault judgment method of distribution network based on multisource information proposed in this paper has been applied to the research and development of ‘distribution network fault location and judgment system under the condition of multiple incomplete information’.