Research On Fault Section Location Methods Based On State Estimation For Distribution Netweorks

With the rapid adoption of distributed generation(DG),the original unidirectional power flow characteristics of distribution networks have been changed.Once a fault occurs,the fault characteristic is different from it in the traditional distribution network,which can easily cause the failure of traditional fault detection and location methods.In view of the current situation of insufficient real-time measurement in distribution network,it is necessary to carry out research on distributed parallel algorithms for state estimation and fault location technology to achieve rapid and accurate self-perception and self-diagnosis of the distribution network,which is conducive to ensuring the safety and reliability of the distribution system,as well as enhancing the ability to absorb renewable energy.Driven by this motivation,on the basis of summarizing the research status on large-scale distribution network partitioning,distributed state estimation,fault location technology,and analyzing the shortcomings of the current methods of distributed generation access to distribution network,the fast state sensing and fault diagnosis technology of large-scale distribution network have been established in this dissertation.The main contributions are stated as follows:1)An optimal partitioning method for distribution networks based on an improved community discovery algorithm is proposed.By abstracting the nodes,branches and electrical distances into vertices,edges and weights in graph theory,an equivalent weighted undirected network is constructed,and the Louvain community discovery algorithm is used to predict the modularity to obtain the community structure of the distribution network under global optimization.Aiming at the inherent over-zoning defect of the community discovery algorithm,the community structure is further optimized with to realize the scale balance of sub-regions,as well as considering constraints such as connectivity within sub-regions and complementarity between sub-regions,so that the overlapping zoning scheme of large-scale distribution network with the characteristics of high coupling within the region and sparse connection between regions is determined.2)A distributed state estimation model of distribution network considering the power characteristics of boundary nodes is established,including local estimation and system coordination process.In the local state estimation model,the multi-source measurement transformation technology is used to integrate pseudo-measurement and real-time measurements such as SCADA and μPMU.Based on the linear iterative method combined with sparse matrix technology,the proposed local state estimation model can obtain the estimated value of the three-phase voltage amplitude and phase angle quickly.In the process of system coordination,the equivalent load model is used to comprehensively consider the sub-regions’ downstream power characteristics and overlap node voltage information to speed up the global convergence process.The distributed state estimation algorithm alternates between local estimation parallel operations and system coordination until satisfying the convergence conditions.3)An improved state estimation method and an implementation strategy for the location of fault sections in distribution networks are proposed.The conventional distributed state estimation is improved using pseudo-measurement,weight matrix correction and the virtual fault equivalent model,and the fault equivalent injected power is accurately calculated.On this basis,a two-step strategy is proposed.Firstly,the faulted sub-area is determined according to the ratio of the estimated errors of the normal state estimation of each sub-area before and after the fault,and then the improved state estimation is used to search the line to determine the location of the single-phase ground faulty section within a small range of the sub-area.