The Study Of Parallel Simulation For Power System Transient Stability

Heretofore, time domain simulation is the most reliable method for power system transient stability analysis. The On-line real-time or the over real-time time domain simulation for the large-scale power system, providing a favorable and well-knit base for the system to operate securely and stably, but still could not be realized because of the huge computational quantity of electromechanical transient stability at present time. However, with the development of distributed parallel processing technology, especially the emergence of extensible and high cost performance PC-Cluster platform brings a hope for this problem to be solved.The main work of this thesis is as follows:Firstly, the simulation method of Power systems transient stability which based on Preconditioned Conjugate Gradient (PCG) is studied. A new algorithm which combines direct method and PCG method is presented. The method adopts incomplete LU factorization to precondition on the jacobian matrix in order to reduce the computation quantity; and then utilizes the characteristic of Transient Stability simulation which in several simulation step the system state variables changs little, to reuse the jacobian matrix and its preconditioned result of the pre-step for subsequent simulation step; and further combines with direct method to speed up the iterative solving process. And the computation quantities of the algorithm and the direct method are done and compared.And then the basic concept of MPI is introduced in brief. Two communication types of MPI namely point to point communication and collective communication which are adopted in this thesis are described in detail.At last, a new parallel strategy and algorithm for power system transient stability analysis based on waveform relaxation method is presented. The process is: the network is divided into several subsystems based on its geography subareas; the neighbour subsystems foreach subsystem are being equivalence processed to the tie lines and boundary buses; further, the equivalence currents transferred to the boundary buses of the relevant subsystems are considered to be the communication waveform information between the subsystems, and the boundary buses are treated as the voltage source; then the subsystems are naturally decoupled from each other, and the parallel calculation is finally realized. The result of the practical sample on the PC-Cluster system demonstrates that the new algorithm possesses high calculation efficiency and parallel speedup ratio, and can achieve on-line calculation for large-scale systems without any speedup measure.