一种电力系统暂态稳定并行计算的优化分区策略

AN OPTIMAL PARTITION SCHEME OF PARALLEL COMPUTING POWER SYSTEM TRANSIENT STABILITY

大规模电力系统的区域特性为暂态稳定并行计算任务的优化划分创造了条件。基于分层递归二分法的思想 ,分析了基于块对角加边策略的空间暂态稳定并行计算的任务划分目标 ,提出了一种基于区域特性的任务划分策略 ,将电力系统自然的分层分区特性与并行计算领域的图划分算法相结合 ,有效地降低了暂态稳定计算中协调系统规模和算法通信量 ,提高了各个计算进程的负载平衡度和并行计算性能。实际大规模电力系统的算例仿真结果表明 ,文中提出的方法优于METIS图划分软件 ,与其划分后计算性能的对比可见 ,暂态稳定并行计算最佳仿真速率提高超过 10 % ,有效地提高了并行计算性能 ,为大规模电力系统实时暂态稳定并行计算的实现奠定了基础。

Dynamics of large scale electric power systems often exhibit regionally which lays the foundation of optimal task scheduling in the parallel computing. In this paper, based on a thought of multilevel recursive bisection and regional features of power network, the goal of task partition for transient stability computation by the parallel algorithm with hierarchical block bordered diagonal form is analyzed. And a new partition scheme is proposed. This scheme combines the natural layered/regional features of power systems with the graphic partition algorithm widely used in the field of high performance computing, and efficiently reduces the border system computation scale and algorithmic communication in transient stability computing environment. Hence the load balancing of each computing process and the overall parallel computing performance are improved. In the simulation cases of a real large scale power system, the scheme proposed in this paper obtains a better partition result than the METIS graphic partition tool and reduces more than 10% of the time consumed in parallel computing. The analysis demonstrates that this new proposed partition scheme can provide a good environment for the real time transient stability parallel computing for large scale power systems.