基于Adomian分解方法的暂态稳定并行仿真研究

Parallel transient stability simulation based on adomian decomposition method

针对目前大规模电力系统难以实现快速实时仿真的问题,提出了一种基于Adomian分解方法的暂态稳定仿真并行算法。首先,在考虑节点权重的同时,采用METIS将大系统拆分成若干子系统,然后采用波形松弛方法对系统进行并行求解。为加速各子系统的迭代过程,所有状态变量经过隐式梯形积分格式差分化后,采用基于Adomian分解的迭代算法配合非诚实牛顿算法进行了求解;为了进一步提高波形松弛法的整体收敛性,同时还使用了窗口方法、预处理方法与波形预测方法;最后,采用2 383节点和12685节点两个算例进行了测试,发电机采用复杂模型,同时考虑励磁调速系统,并通过共享内存的并行环境加以实现。测试研究结果表明,上述算法可以取得较为理想的收敛速度和并行加速比,同时实现了上万节点的超实时仿真。

Aiming at the problem that fast simulation for large-scale power systems is hard to realize,a parallel transient stability simulation based on Adomian decomposition method was presented.First of all,a power system described by large-scale differential-algebraic equations （DAE） was decomposed into several subsystems by METIS with the weight of all nodes taken into consideration.Waveform relaxation method was adopted for parallel implementation and all subsystems were solved using Adomian decomposition method combined with very dishonest newton （VDHN） method independently after the state variables were discretized by implicit-trapezoidal rule.Moreover,to accelerate convergence of the system,windowing,preconditioning and waveform prediction were adopted.Finally,the proposed algorithm was verified by two large-scale test cases with 2 383 buses and 12 685 buses and complex model was adopted for generators as well as the exciting and governing systems.The parallel efficiency was further improved by a shared-memory parallel environment.The results indicate that the proposed method achieves in increasing convergence speed and parallel speedup effectively,and super real-time simulation is realized.