Research On Parallel Waveform Relaxation Method For Large-Scale Power System Transient Stability Simulation

As the scale of the power system grows, it becomes ever more important to increase the reliability of the power grid require, which requires fast and accurate simulations to properly analyze power system transient stability. In the last few decades, the developments in computer technology, particularly parallel computing, has provided a solid foundation for realizing real-time, or even super real-time simulations of transient stability analysis for large-scale power systems. As a result, it is quite crucial to find a proper parallel algorithm for transient stability simulation for large-scale power systems.In this paper, researches on parallel waveform relaxation method for large-scale power system transient stability simulation are presented.(1) Waveform relaxation method is introduced and the implementation for transient stability simulation is proposed. The models of dynamic units which are adopted in this paper are introduced. After that, the basic process of waveform relaxation and the corresponding iteration format is presented. To accelerate convergence of the system, the method adopts a windowing approach, preconditioning, and waveform prediction. Finally, two large-scale test cases with2383nodes and12685nodes verify the proposed algorithm. The result shows that the waveform relaxation method can be applied for power systems with tens of thousands buses with required accuracy. And the preconditioning and waveform prediction approach can accelerate the convergence of effectively. And the serial simulation in the paper is faster than the implicit trapezoidal integration method.(2) Iteration methods based on Adomian decomposition method are adopted for transient stability simulation. And the proposed method was used to solve the subsystems for waveform relaxation method. Adomian polynomial is introduced and Adomian series are derived correspondingly. Then different forms of Adomian series are presented, and several iteration formats based on different Adomian decomposition methods are tested for transient simulation. The results show that the iteration method based on Adomian decomposition is more efficient than the Newton-Raphson method. The calculation brought by the extra step is much less than the simulation time which is saved by the method. And the Adomian decomposition method can not only be used for waveform relaxation method, but also be applied for transient simulation.(3) A parallel-in-space and parallel-in-time strategy is presented for parallel implementation in this paper. First, the system is decomposed into several subsystems by a partitioning algorithm based on epsilon decomposition. Besides, to accelerate the parallel speedup, a two-state parallel strategy is proposed which makes the algorithm totally paralleled. Moreover two kinds of parallel programming models are used for the implementation of the algorithm in this paper to test the parallel speedup. The results show that the partitioning strategy in this paper is more appropriate with better convergence speed. And a faster-than-real-time simulation is realized after the parallel strategy being adopted. Compared with MPI, OpenMP performs better for the fine-grained parallel algorithm. The max speedup in this paper is10.38with a proper convergence speed, which validate the efficiency of the proposed parallel algorithm.