Research And Application Of Dynamic Partitioning For Large Power Grid Setting Calculation

With the development of region network interconnected and the grid scaleunceasingly expanded, the traditional partition structure is becoming more and moreblurred and each unit at all levels influence each other. Large power grid unifiedsetting calculation based on the integrated application has become the developmentdirection with this background. However, the efficiency of integrated settingcalculation is low because of the large grid size. Block Parallel computing is animportant way to improve computational efficiency. In order to avoid theunreasonable traditional block structure and the defects of boundary slip settingcalculation, this paper proposes a method that the network size in each computerconsists of the setting calculation region and buffer subsystem. Each computercompletes all elements setting calculation in its setting calculation region. Meanwhilebuffer subsystem can improve the reliability of boundaries and their nearby branches.In order to build an integrated setting calculation model, this paper analyzes thedesign concept of base data platform, which consists of horizontal integration andvertical run-through. On this basis, we use the method of data splicing to buildintegrated setting calculation model.In order to improve the efficiency of parallel computing, the setting calculationworkload of each computer must be balanced. An optimal partition method based onthe virtual impedance matrix is used to determine the setting calculation region ofeach computer. This method comes from the complex network theory. There areoptimal community structures in different number of partitions. They are obtained byusing virtual impedance matrix and boundary nodes migration. Under these, theshortest calculation time is used as the objective function. Meanwhile, the optimalresult is obtained by comparing different partition results.The large power grid setting calculation dynamic partitioning based on buffersubsystem is put forward. The purpose of dynamic partitioning of large power grid isto obtain the network scale in each computer. The buffer subsystem for settingcalculation consists of the coordination region and constraint region. Afterdetermining the scope of setting calculation, we include the adjacent branches ofboundaries and its end nodes in the coordination region according to remote backupprotection. Furthermore, the shortest electrical path is used to determine theminimum ring network of boundaries and coordination branches. Meanwhile thesubstation scope is determined by the self-impedance variation rate of boundarynodes. The constraint region consists of the smallest ring structures and substationscope. Under these, the depth first search and backtracking technique are applied to realize the dynamic partitioning of large power grid automatically.