Parallel Mpi-based Power System Load Flow Computation

The modern electric power system has become a special system that contains the attribute of large capacity, super high voltage, wide area, and complex structure. The power flow calculation is one the fundamental calculations performed for the operation of the whole power system. However, the traditional computation methods are incapable of dealing with real-time analysis with huge amount of data. The development of power system operation calls for faster processing and computation. Simultaneously, as the continuing development of computing technology, and emergence of high-performance cost efficient computers and high-speed network, the parallel computers are produced and put into practical use. The introduction of parallel processing into electric power system computation provides an efficient way to solve a series of problems. As the electric power system becomes larger and larger, more and more factors need to be considered, which resulted in a huge amount of equations in calculation. The parallel power flow calculation is becoming an important method to enhance the speed of time-consuming multiple data power flow operation. And there exists low-cost and high-efficiency solution if the parallel platform is constructed on a cluster of PCs. This thesis first introduces the basic theory of parallel computing, the fundamentals of parallel computers and parallel algorithms. Them the MPI (Message Passing Interface) for PC is introduced, and an experimental setup based on Windows XP and PC cluster is built. There is distinguished difference between MPI programming and traditional methods. The MPI programming on PC cluster is discussed and a power flow calculation program is implemented. Two test system, IEEE 14 and IEEE 118 bus systems are used to testify the program and to measure the performance. The thesis further discussed the effect of the network communication (message passing) on the parallel computing efficiency. Finally, based on the study and experimental results, it is concluded that it is feasible to use PC cluster to perform power flow calculation.