摘要
以提升大规模组合故障快速仿真分析能力为目标,在Hadoop框架下研发了连锁故障分布式计算技术。基于PSD-BPA软件计算模块,利用Java开发连锁故障计算分析功能,实现驱动判定、故障集筛选、事故链搜索、严重度评估4类模块。通过部署Hadoop分布式文件系统(HDFS)存储调度功能,将事故链解耦为小粒度单一故障场景进行计算,可针对连锁故障仿真的不同复杂度提供跨系统的分布式计算服务,灵活应对计算开始前连锁故障中事故链组合的不可预测性。利用10机、16机系统和某省网实际数据进行技术测试,结果表明所研发系统实现了连锁故障分析应用与数据在计算服务网络中的分离,具备动态调配计算节点资源的能力,能自动适应事件规模为电网连锁故障的仿真分析提供强大计算能力,具有在线应用前景。
Aiming at improving the computing capability of power system cascading failure analysis,this paper proposes a platform which employs a distributed computing framework,Hadoop as an underlying infrastructure for power system cascading failures simulation and evaluation.Based on the power flow computation and time-domain simulation supported by PSD-BPA suites,the research significantly functions the action tripping logic,the filtering of pre-defined contingencies,fault chains searching and severity measurement.By means of the deployment of Hadoop framework with its distributed file system and customized dispatching program,our platform can integrate multiple simulation modules and supply services in processing computational intensive cascading failure scenarios.The experiments were carried out using the data of two benchmark systems and one practical power system respectively.The results indicate that the platform can dispatch computing tasks to nodes with considering load balancing.And it is self-adaptively processing the scale of fault chains in parallel.It is proved advanced stability and security analysis for power system.The proposed framework has the potential for on-line.
出处
《电力系统自动化》
EI
CSCD
北大核心
2016年第7期90-97,共8页
Automation of Electric Power Systems
基金
国家自然科学基金资助项目(51207098)~~
