基于Beowulf集群的大规模电力系统方程并行PCG求解  

Parallel PCG Solution of Large Scale Power System Equations Based on Beowulf Cluster

作　　者：刘洋[1] 周家启[1] 谢开贵[1] 赵渊[1] 陈炜俊[1] 胡博[1] 

机构地区：[1]重庆大学高电压与电工新技术教育部重点实验室,重庆400044

出　　处：《电工技术学报》2006年第3期105-111,共7页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金(50307015);重庆市科技计划(2003-7951)资助项目。

摘　　要：研究基于Beowulf集群的大规模电力系统方程并行PCG求解问题,采用常见的硬件设备和廉价且广为传播的软件构建出Beowulf分布式集群环境,基于这一环境,彻底摒弃矩阵的传统直接分解算法,采用多项式预处理的PCG法并行求解大型稀疏线性方程组。文中方法无需进行任何形式的电网络划分,也无需进行任何的矩阵三角分解和前推回代过程,适合各种类型的大规模电力系统方程的并行求解。本文分别在潮流计算、状态估计和静态安全分析中对大规模电力系统方程实现了并行求解,并获得一定的加速比和并行效率,为大规模电力系统快速、准确的仿真计算和在线分析提供一种可行的新途径。It presents a detailed investigation of the parallel PCG method for large-scale power system equations based on Beowulf cluster. With common hardware and freely available software, a clustered PC computing system named Beowulf cluster for power system is given. And based on the distributed platform, the PCG method is used instead of the traditional LU direct method for solution of the sparse sets of linear equations of large-scale power system. The proposed method based polynomial preconditioning is easy to parallelize due to avoid factorization of the large matrices. Tests were performed with networks of several synthesized bulk systems related to power flow calculation, state estimation, and static security analysis, and showed some speed-up ratios and parallel efficiencies.

关 键 词：BEOWULF集群 并行PCG法 潮流计算 状态估计 静态安全分析 

分 类 号：TM744[电气工程—电力系统及自动化]