| 引用本文: | 吴云芸,方家琨,艾小猛,薛熙臻,胡伟,沈煜,文劲宇.计及多种储能协调运行的数据中心实时能量管理[J].电力自动化设备,2021,41(10): |
| WU Yunyun,FANG Jiakun,AI Xiaomeng,XUE Xizhen,HU Wei,SHEN Yu,WEN Jinyu.Real-time energy management of data center considering coordinated operation of multiple types of energy storage[J].Electric Power Automation Equipment,2021,41(10): |
|
| 摘要: |
| 随着互联网+、云计算的发展，数据中心能耗迅速增加，高能耗和高电费问题日益突出，对数据中心进行能量管理和优化是运营商提升市场竞争力的重要手段。但由于数据负荷、电网电价和新能源出力的不确定性，如何在实时运行时保证数据中心的运行经济性是亟待解决的问题。针对以上问题，考虑数据负荷调度、服务器休眠、多种储能协调运行、与电网交互等因素，建立了数据中心的实时能量管理模型。由于模型中多类型储能和批处理负荷各自的时段间耦合约束都会影响系统全局最优决策，需要分别对其进行解耦，故提出一种基于多维分段线性函数近似值函数的近似动态规划(PLF-ADP)算法的数据中心实时能量管理策略。仿真算例表明，所提多维PLF-ADP算法能够在随机环境下考虑数据中心中多类型储能和批处理负荷的协调运行，得到近似全局最优的实时能量管理策略，保证数据中心运行的经济性。 |
| 关键词: 数据中心 实时能量管理 近似动态规划 批处理负荷 储能协调运行 马尔科夫决策过程 |
| DOI：10.16081/j.epae.202110016 |
| 分类号:TM73 |
| 基金项目:湖北省技术创新重大专项(2019AAA015) |
|
| Real-time energy management of data center considering coordinated operation of multiple types of energy storage |
|
WU Yunyun1, FANG Jiakun1, AI Xiaomeng1, XUE Xizhen1, HU Wei2, SHEN Yu2, WEN Jinyu1
|
|
1.State Key Laboratory of Advanced Electromagnetic Engineering and Technology, Huazhong University of Science and Technology, Wuhan 430074, China;2.State Grid Hubei Electric Power Research Institute, Wuhan 430077, China
|
| Abstract: |
| With the development of “Internet+” and cloud computing, the energy consumption of data centers is increasing rapidly. The problems of high energy consumption and high electricity bills are becoming increasingly prominent. Energy management and optimization of data centers are important means for operators to enhance their market competitiveness. However, due to the uncertainty of service requests, electricity price and renewable energy output, how to ensure the economic real-time operation of the data center is an urgent problem to be solved. In response to the above problems, a real-time energy management model for the data center is built, which considers factors such as job scheduling, server sleep policy, coordinated operation of multiple types of energy storage, and interaction with the power grid. Since the coupling constraints among the intervals, which exist in both energy storage system and batch jobs, will affect the global optimal decision of the system, they needs to be decoupled separately. Therefore, a data center real-time energy management strategy based on multi-dimensional PLF-ADP(Piecewise Linear Function based Approximate Dynamic Programming) algorithm is proposed. Simulation examples show that the proposed multi-dimensional PLF-ADP algorithm can consider the coordinated operation of multiple energy storage and batch jobs in the data center under uncertain environments, and obtain the approximate global optimal real-time energy management strategy to ensure the economy of data center operation. |
| Key words: data center real-time energy management approximate dynamic programming batch jobs coordinated operation of energy storage Markov decision process |
