智能电网环境下分布式因特网数据中心的能量管理研究

发布时间：2018-04-25 17:27

本文选题：智能电网 + 分布式因特网数据中心　；参考：《华中科技大学》2014年博士论文

【摘要】：随着云计算和各种因特网在线业务的迅猛发展,越来越多的数据和计算被迁移至分布在不同地理位置上的因特网数据中心。因特网数据中心有许多优点,譬如：数据可靠、便于客户管理、摊销成本低。然而,在因特网数据中心的运行中存在几个主要问题,譬如：能量消耗和能量成本巨大,碳排放量较高。为了降低能量消耗、能量成本以及碳排放量,需要有效地管理因特网数据中心的能量消耗方式和能量供给方式。考虑到智能电网的巨大潜力,本文研究了智能电网环境下分布式因特网数据中心的能量管理问题。首先,本文研究了基于电力中断约束的分布式因特网数据中心能量管理问题。在未来电价、负荷、可再生能源输出及大电网运行状态不确定性环境下,我们建模了时间平均期望能量成本最小化问题并设计了一种求解算法,且该算法无需知晓任何系统参数的统计信息。理论分析了该算法的可行性及其能提供的性能保证。基于实际数据的仿真表明了所提算法的有效性。其次,本文研究了基于碳排放约束的分布式因特网数据中心能量管理问题。在未来电价、负荷、可再生能源输出及碳排放率不确定性环境下,我们建模了能量成本与碳排放加权和的时间平均期望值最小化问题并设计了一种求解算法,且该算法无需知晓任何系统参数的统计信息。理论分析了该算法的可行性及其能提供的性能保证。基于实际数据的仿真表明：与现有的碳中立算法相比,所提算法可同时取得更低的能量成本和碳排放量。再者,本文研究了基于运行风险约束的分布式因特网数据中心能量管理问题。在提出风险受限的决策框架后,我们在未来现货电价和负荷不确定性环境下建模了风险受限的期望能量成本最小化问题。为了求解优化问题,我们提出了基于分解的切平面算法。基于实际数据的仿真表明：所提决策框架可根据因特网数据中心运营商的风险偏好实现期望能量成本和运行风险之间的最优折中。最后,对全文工作进行了简要的总结,并对下一步研究工作进行了展望。
[Abstract]:With the rapid development of cloud computing and various Internet online services, more and more data and computing are migrated to Internet data centers located in different geographical locations. Internet data centers have many advantages, such as reliable data, easy customer management and low amortization costs. However, there are several major problems in the operation of Internet data centers, such as the huge energy consumption and energy cost, high carbon emissions. In order to reduce energy consumption, energy cost and carbon emissions, it is necessary to effectively manage the energy consumption and energy supply of Internet data centers. Considering the great potential of smart grid, the energy management of distributed Internet data center in smart grid environment is studied in this paper. Firstly, the energy management problem of distributed Internet data center based on power interruption constraint is studied in this paper. In the uncertain environment of future electricity price, load, renewable energy output and the operating state of large power network, we model the time average expected energy cost minimization problem and design a solution algorithm. And the algorithm does not need to know any statistical information of system parameters. The feasibility of the algorithm and its performance guarantee are analyzed theoretically. The simulation based on the actual data shows the effectiveness of the proposed algorithm. Secondly, this paper studies the energy management of distributed Internet data centers based on carbon emission constraints. In the uncertain environment of future electricity price, load, renewable energy output and carbon emission rate, we model the time-average expectation minimization problem of weighted sum of energy cost and carbon emissions and design a solution algorithm. And the algorithm does not need to know any statistical information of system parameters. The feasibility of the algorithm and its performance guarantee are analyzed theoretically. The simulation based on the actual data shows that compared with the existing carbon-neutral algorithm, the proposed algorithm can achieve lower energy cost and lower carbon emissions simultaneously. Furthermore, this paper studies the energy management problem of distributed Internet data centers based on running risk constraints. After putting forward a risk-constrained decision framework, we model the expected energy cost minimization problem with risk limitation in the future spot electricity price and load uncertainty environment. In order to solve the optimization problem, we propose a decomposition-based tangent plane algorithm. Simulation based on actual data shows that the proposed decision framework can achieve the optimal compromise between expected energy cost and operational risk according to the risk preference of Internet data center operators. Finally, a brief summary of the full-text work is given, and the future research work is prospected.
【学位授予单位】：华中科技大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TM769

【参考文献】