摘要
针对目前电力系统仿真计算实时性与计算精度要求高、平台可扩展性差以及资源利用率低等特点,首先给出了一种基于开源基础设施平台OpenStack和并行处理框架Hadoop的电力仿真云计算平台架构,能够以较低成本实现动态扩展、高效计算和海量存储等功能。其次,结合电力系统仿真任务特点,给出了一种基于多目标粒子群优化(PSO)算法的虚拟机迁移策略,实现电力仿真云计算平台资源调度。虚拟机迁移过程采用指数平滑预测模型确定热点,选择虚拟机时综合考虑迁移速度和效果两个因素,利用多目标PSO算法搜索目标节点,使得电力系统仿真计算在保证服务质量的同时兼顾高资源利用率和低运行成本的优势。最后,通过CloudSim进行仿真实验,将所提算法与贪心迁移算法和顺序放置非迁移算法进行对比。实验表明,所提算法在服务等级协议(SLA)违背率、剩余资源率、能耗以及虚拟机迁移次数等指标上均优于其他算法,验证了基于虚拟机动态迁移的多目标PSO算法在电力仿真云计算平台资源调度中的优势和可行性。
In view of the requirements of high real-time and calculation accuracy,poor platform scalability,and low resource utilization of power system simulation calculation,an architecture of cloud computing platform for power system simulation based on the OpenStack(an open source infrastructure platform)and Hadoop(aparallel processing framework)is presented for achieving dynamic expansion,efficient computing and massive storage functions at a low cost.Then,by referring to the characteristics of power system simulation task,a virtual machine migration strategy based on multi-objective particle swarm optimization(PSO)algorithm is given for achieving resource scheduling of the cloud computing platform.During the virtual machine migration process,the predictive model based on exponential smoothing is used to determine the hotspots,and the virtual machines are selected with the migration speed and effect taken into account.Meanwhile the multi-objective PSO algorithm is used to search for the target node to ensure the quality of service of power system simulation while attending to the advantage of high resource utilization and low operating costs.Finally,simulation experiments are conducted on the CloudSim to compare the algorithm proposed with the greedy algorithm with migration and the non-migratory algorithm with sequence placement.The results show that the algorithm proposed is batter than the other two in terms of the rate of service level agreement(SLA)violation,the rate of surplus resources,energy consumption and the times of virtual machine migration.The advantages and feasibility of multi-objective PSO algorithm based on virtual machine migration in the resource scheduling of cloud computing platform for power system simulation are verified by experiments.
出处
《电力系统自动化》
EI
CSCD
北大核心
2015年第12期97-105,共9页
Automation of Electric Power Systems
基金
国家自然科学基金资助项目(61074078)
中央高校基本科研业务费专项资金资助项目(12MS113)~~
关键词
云计算
虚拟机
电力系统仿真
粒子群优化算法
动态迁移
cloud computing
virtual machine
power system simulation
particle swarm optimization(PSO)algorithm
dynamic migration
