【作者】 吴刚；

【导师】 高赐威； 周珏；

【作者基本信息】 东南大学 ， 电气工程（专业学位）， 2019， 硕士

【摘要】 近年来,数据中心能耗增长迅速,其电力消耗已具有相当体量且仍呈上升趋势。与传统需求响应资源不同,数据中心负荷同时具有时间、空间维度上的可转移特性,该特性使之越发受到国内外研究关注。然而,考虑数据中心负荷的电网规划及运行理论尚不完善,电网侧尚未充分意识到地理上分散的数据中心在降低电能传输损耗、缓解跨区域电网阻塞、提高可用输电能力等方面的利用价值,亟需开展相关研究。本文从结合需求响应分析数据中心用电负荷优化出发,对考虑负荷转移特性的数据中心参与电网规划运行进行了研究。在运行层面,考虑数据中心负荷时间维度转移特性,研究了平抑风电波动策略;考虑空间维度转移特性,研究了电网降损以及阻塞管理方法。在规划层面提出考虑负荷空间维度转移特性的电网规划流程,并探究其对规划方案中系统可用输电能力的影响。首先,概述了国内外数据中心的发展现状,调研了数据中心能耗建模方法,分别概括了包括数据中心在内的需求侧资源平衡风电功率波动、参与电网经济运行、参与电网规划等方面的研究现状,为后续研究奠定基础。其次,分析了数据中心的负荷优化方式;研究了不同管理模式的数据中心参与需求响应的方式;从直接负荷控制、电价信号引导、参与电力市场3个方面讨论了数据中心需求响应的具体形式。再次,研究利用数据中心时间维度上负荷可转移特性平抑短时间风电波动。建立数据中心负荷调节量与风电平抑量最优匹配模型,分析不同提前通知时长对实际平抑效果、数据中心功率变化程度的影响。求解二次规划问题得出的算例结果表明,优化各时段与可延迟网络负载对应的功率,能够有效平抑风电短时波动。然后,研究考虑负荷空间转移特性的数据中心参与电网经济运行。建立考虑服务器CPU频率调节的电网降损模型,分析网络负载响应时限及网络负载量对网损下降趋势的影响;建立给予数据中心激励的阻塞管理模型,提出计及数据中心跨区域负荷调节能力的阻塞管理流程。基于IEEE30节点系统和5区域互联系统的算例结果表明,数据中心间网络负载的优化分配,有利于电网实现降低网损、缓解联络线阻塞的目标。最后,研究利用数据中心负荷空间转移特性的电网规划。建立计及负荷空间转移的可用输电能力修正模型,提出考虑数据中心负荷转移特性的电网规划流程。以18节点系统进行算例分析,运用模糊集对分析方法比较待选方案的优劣,结合具体方案中网络拓扑下的ATC瓶颈支路及负荷转移可行域,分析了数据中心负荷转移对提高方案灵活性指标的影响。研究结果表明,在规划阶段考虑可空间转移的数据中心负荷,能够有效应对规划方案中的网络输电瓶颈,提高规划方案的灵活性。更多还原

【Abstract】 In recent years,energy consumption of data centers has grown rapidly,and the consumption scale has been quite large and still rising.Different from the traditional demand response resources,the data center load has the transferable characteristics in both time and space dimensions,which makes it more and more concerned by domestic and foreign research.However,the grid planning and operation theory considering data center load is still not perfect,and the grid side is not fully aware of the utilization value of geographically distributed data centers in reducing power transmission loss,easing cross-regional grid congestion,and improving available transmission capacity(ATC).There is an urgent need to conduct relevant research.This thesis studies data center’s participation in the power grid planning and operation considering the load transfer characteristics,which starts from the analysis of data center power optimization with the issue of demand response.At the operation level,considering the time-scale load-shifting characteristic of data center,the strategy of smoothing wind power fluctuation is studied.Besides,the grid loss reduction and congestion management methods are studied considering the spatially transferable characteristic of data center load,and the grid planning process considering this characteristic is proposed at the planning level.Furthermore,the impact of the spatially transferable characteristic on the system’s ATC is explored.Firstly,the development status of data centers at home and abroad is summarized,and the data center energy consumption modeling methods are investigated.The status of researches on smoothing wind power fluctuations,power system economic operation,and power grid planning considering demand side resources including data center is summarized,laying the foundation for following research.Secondly,the thesis analyzes the load optimization method of the data center,and the ways in which data centers of different management modes participate in demand response are researched.The specific demand response forms of the data center are discussed from three aspects: direct load control,electricity price signal guidance and participation in the power market.Thirdly,the study uses the time-scale load-shifting characteristic of the data center to smooth short-term wind power fluctuations.The optimal matching model of data center load regulation and smoothing capacity is established,and the influence of different advance notice durations on the actual effect and the degree of data center power change is analyzed.The results of the example obtained by solving the quadratic programming problem show that optimizing the power related to the delay-tolerant workload in each period can effectively smooth the short-term fluctuation of wind power.Then,the data center’s participation in the economic operation of the grid considering the load transfer characteristic is studied.A grid loss reduction model considering the CPU frequency adjustment of the data center server is established,and the impact of workload response time limit and workload amount on the loss downtrend is analyzed.Besides,a congestion management model that considers the incentives given to data centers is established,and a congestion management process that takes into account the cross-regional load adjustment capability of data centers is proposed.The results of the examples based on the IEEE-30 system and the 5-area interconnected system show that the optimal allocation of workload among data centers is beneficial to the grid to achieve the goal of reducing network loss and easing tie line congestion.Finally,the thesis studies the grid planning utilizing the spatially transferable characteristic of data center load.An ATC correction model that takes into account load transfer is established,and a grid planning process is proposed.The 18-node system is used for case study,and the fuzzy set pair analysis method is used to evaluate the candidate scheme.Based on the ATC bottleneck branch and the load transfer feasible domain under the network topology,the impact of load transfer on the flexibility improvement of the scheme is analyzed.The research results show that if the data center load that can be spatially transferred is considered in the planning stage,it can effectively cope with the network transmission bottleneck in the planning scheme and improve the its flexibility.更多还原