摘要
为实现大规模配电网的实时分析计算,构建了基于数据级任务分解的配电网分布式并行计算平台。结合配电网运行结构和设备配置,以配电网馈线作为分析单元,采用数据级并行计算方式对配电网计算任务进行分解。配置管理模块、实例、执行端和客户端4个子系统,分别实现任务生成、任务分解、任务派发以及子任务计算等功能,形成分布式并行计算平台框架。引入消息中间件ZeroMQ技术,采用不同类型套接字的组合实现分布式系统内部的N-N高效通信及与外部系统的数据交互。为验证该平台的实用性及并行计算性能,在平台上实现了山东省某市的城市配电网全局状态估计的分布式并行计算。算例分析表明:在配电网节点达到一定规模时,采用该平台进行分布式并行计算具有明显的速度优势。
To implement real-time analysis and computation for large-scale distribution system the data-level task decomposition based distributed and parallel computing platform for distribution system is constructed. Combining with operational structure and equipment configuration of distribution system, taking distribution feeder as analytical unit and adopting data-level parallel computing mode the computation task of distribution system is decomposed. Four subsystems, namely management module, living examples, execution end and client, are configured to implement such functions as task generation, task decomposition, task distribution and sub-task computation to form a frame for distributed and parallel computing platform. Leading in message-oriented middleware technology ZeroMQ and adopting the combination of different types of sockets, the data interchange of N-N high efficient communication inside the distributed system with external system is implemented. To verify the practicality and parallel computing performance of this platform, the distributed and parallel computation of global state estimation for an urban distribution network of a certain city in Shandong province is implemented by the proposed platform, and calculation results show that when the numbers of distribution nodes reach a certain size, obvious speed superiority can be achieved by use of the proposed platform.
出处
《电网技术》
EI
CSCD
北大核心
2014年第4期1103-1108,共6页
Power System Technology
基金
中央高校基本科研业务费专项资金资助项目(11MG13)~~
关键词
数据级并行
分布式并行计算平台
配电网
分布式通信架构
ZeroMQ
data-level parallel
distributed and parallel computing platform
distribution network
ZeroMQ
distributed communication architecture
