基于集群机的大规模电力系统暂态过程并行仿真

由于传统的串行计算方法无法满足互联电力系统在线动态安全分析和实时仿真的要求，研究高效的电力系统暂态并行仿真算法及其软件已成为电力系统动态实时仿真的关键。该文基于集群系统，提出一种利用电力系统区域特性的多重化网络划分方案，弥补了传统网络划分算法在计算规模和划分质量上的不足；并提出了一种基于分块思想、消息传递与动态多线程相结合的网络方程分层算法及其实现策略，缩短了网络方程的计算时间，解决了暂态仿真中计算量突增的问题，提高了算法的并行效率和整体性能。同时，文中还简化了暂态稳定算法迭代流程，改进了全局收敛判定方法，有效地减少了算法计算量和通信损耗。实际电网的数值计算结果表明，文中提出的算法计算加速比高，实时性好，因此研制的电力系统动态并行仿真程序能够满足大规模电力系统暂态过程实时仿真、甚至超实时仿真的要求。

CLUSTER-BASED PARALLEL SIMULATION FOR POWER SYSTEM TRANSIENT STABILITY ANALYSIS

With the development of modern power industry, the traditional serial algorithm for transient stability computation cant fulfill the requirements of on-line security analysis and real-time simulation. The research on high-powered parallel transient stability algorithm and its implementation become more and more important. This paper analyzed the advantages and disadvantages of available parallel algorithms and presented a multilevel partition scheme, which was based on power network regional characteristics, to reduce the computation scale and enhance the quality of network partition. To improve the performance and scalability of the parallel simulation, a new hierarchical Block Bordered Diagonal Form (BBDF) power network algorithm was proposed in this paper and mapped to cluster system using message-passing and share-memory models simultaneously. Meanwhile, this paper also simplified the iteration flow of transient stability computation and improved the scheme of convergence checking. Numerical results of actual large-scale power networks have shown that the parallel algorithm on cluster systems can gain high speedups and run much faster than real-time dynamic process. The parallel simulator developed with this algorithm can achieve real-time transient stability simulation for large-scale power grids.