Power Flow Grid Platform Parallel Algorithm To Achieve

The scale of modern power system becomes larger and larger. In order to facilitate the overall simulation of power system, effectively strengthen the stability and security, to avoid large-scale power system failure, it is necessary to help the planners and dispatchers to manage the operation and control of power system with the state-of-art high performance computing that is suitable for power system application. The research in parallel computing in power system was started long ago. Many algorithms are proposed and the theory about centralized parallel computing is mature to some extent. However, as the development of science and technology and the reformation of power system, it is necessary to further improve the efficiency of parallel computing and take some measures to make the parallel computing be used more easily.This thesis proposes to establish a distributed parallel computation platform for the power system, using the novel grid computing and distributed processing technology. This technology can realize the formation of dynamic parallel computing workgroup by organizing the dispersed and heterogeneous computing resources available network. This service is compatible with the information sharing and services available on current network. It makes it possible to form a distributed parallel computing architecture for the inherently distributed power system.The thesis mainly focuses on the parallel solution of power flow steady state analysis application in power system in the distributed grid computing environment. For this large-scale non-linear equation, a blocked parallel algorithm for the power flow Jacobi matrix equation is designed based on the parallel triangular decomposition. Through (multi-level) bordered block partition of the power system topology matrix, a BBDF coefficient matrix which is suitable for above decomposition is formed. When the equation is solved with parallel iterative method, convergence acceleration is achieved through damping Newton method. At last, algorithm is implemented as a grid service and deployed at Globus grid nodes. This service is called by clients to accomplish computing task. The thesis also designed the MPI program for performance comparison purpose.