Parallel computation of power system transient stability based on symplectic Gauss algorithm and preconditioned GMRES method

This paper presents a novel parallel algorithm for power system transient stability computation. The proposed algorithm uses the s-stage 2s-order symplectic Gauss method to convert the differential-algebraic system simultaneously at s time points into a set of nonlinear algebraic equations, and the algebraic system is then solved using Newton's method. By the use of the matrix factorization technique, the solution of the linear equations involved in Newton's process is decomposed into two parts: the first part is decoupled according to the stage or at different time points, thus it is fully parallelizable-in-time, and the second part is solved using a preconditioned GMRES method while a new preconditioning way has been proposed. The convergence of the proposed algorithm has been tested on three example power systems. Furthermore, the proposed algorithm has been implemented on a single GPU based computer, and the results show the proposed algorithm achieves high speed-up ratio, and can be applied to the real-t