The Application Of A Graph Partitioning Algorithm In The Large-Scale Power System

Since the formation of the national power grid, the demands for computation for the on-line simulation of power systems are getting increasingly higher and higher. The traditional serial computation for power flow methods can not fulfill the needs of the power grid.The efficient serial computation for power flow and distributed computation theoretical study is becoming the key to large-scale power systems simulation.Therefore,by studying a kind of general parallel algorithm, efficient computation and dynamic simulation with forecast, real-time monitoring on-line systems can be achieved. It is of great significance to modernize the operation and control of power systems.Network partitioning giving rise to the bordered block diagonal form (BBDF) admittance matrix is an effective approach in the field of power system parallel computation. A graph partitioning algorithm based on geographic information of power system was presented for the purpose of transforming the coefficient matrix of equations into nested BBDF (NBBDF), via edge cut set in a hierarchical way. Without destroying BBDF graphs'good performance in each layer, distributed storage of power grid data is adopted in this algorithm, which has little time overheads. In order to decrease fill-ins in the computation, it is necessary to renumber those nested borders produced by this algorithm through sparse technique.This article first introduces the basic methods of computation for power flow and state estimation. Then it elaborates the basic methods of power systems with graphs. It systematically describes the relationship nested with BBDF graphical theory and figure process. This algorithm will be validated through the eastern 2806 node system and the IEEE300 system. Then the algorithm will be introduced to serial computation for power flow and state estimation, which gives a detailed calculation process. This algorithm which is based on the MPI parallel computation environment actualizes simulation calculation. The result proves the proposed algorithm has smaller traffic than any other BBDF decomposition algorithm. These results show that this algorithm which can be carried out in large series of computations for power flow have strong adaptability in geographical regions.