The Research For Parallel Computing Of Transient Stability Based On Symplectic Gauss Algorithm

The analysis of power system transient stability has become more and more difficultalong with the power grid scale increasing, it has been difficult to achieve fast accuratecalculation and analysis for it by traditional calculate method and the serial simulation of asingle computer. Therefore, we hope a new calculation method and parallel computingarchitecture to complete this task.In this article, we have a deep study on the calculation and analysis of large scalepower system transient stability depend on National Natural Science Foundation ofChina(50977052) and Guangdong Grid Corporation Technology Project(DK0010D-T0002). We use s-class2s-order symplectic Gauss method for power system transientstability computing, and put forward two effective and practicable parallel computingmethods. The proposed algorithm uses the s-stage2s-order symplectic Gauss method toconvert the differential-algebraic system simultaneously at s time points into a set ofnonlinear algebraic equations, and the algebraic system is then solved using Newton’smethod. By the use of the matrix factorization technique, the solution of the linearequations involved in Newton’s process is decomposed into two parts: the first is fullyparallelizable-in-time, and the second is solved using a preconditioned GMRES methodwhile an efficient preconditioner has been proposed for iterative method. Also, we putforward two different kinds of preconditioned methods for it, one is combining multifrontalmethod and GMRES (Generalized Minimal Residual method); another is based onW-trasformation in symplectic geometry and matrix splitting preconditioned method. Boththe algorithms have good convergence and high accuracy. In addition, we use the GPUprogramming techniques to realize the proposed algorithms in the text. For test, theconvergence of the proposed algorithm has been examined on3example power systems,IEEE145bus system, IEEE300bus system and case2383wp bus system, on theGPU-CPU based computer, and the results show the proposed algorithm achieves greatcomputational efficiency relative to the traditional CPU computing.