PDF(4905 KB)
基于两阶段拉格朗日松弛-交替方向乘子法的多数据中心算电协同优化
孙亮, 杨世龙, 党翠, 张宗振, 曹鑫雨, 江守其
电力建设 ›› 2026, Vol. 47 ›› Issue (11) : 187-208.
PDF(4905 KB)
PDF(4905 KB)
基于两阶段拉格朗日松弛-交替方向乘子法的多数据中心算电协同优化
Two-Stage Lagrangian Relaxation-Alternating Direction Method of Multipliers Based Computing-Power Coordinated Optimization for Multiple Data Centers
【目的】随着“东数西算”推进，多数据中心（data centers，DC）用电需求与新能源出力的时空不匹配问题日益突出。针对现有研究对数据中心功率映射刻画不足、算力侧与电力侧协调机制不够完善的问题，本文提出一种基于两阶段拉格朗日松弛-交替方向乘子法（Lagrangian relaxation-alternating direction method of multipliers， LR-ADMM）的多数据中心工作负载时空迁移算电协同优化方法。【方法】首先，建立计及信息技术（information technology，IT）设备、辅助设备和制冷系统耦合关系的精细化数据中心功率模型，并在工作负载时空迁移模型中引入服务质量、网络时延和链路带宽约束。其次，构建以购电成本、碳成本、弃风弃光成本、迁移成本和服务质量损失成本最小为目标的协同优化模型。然后，设计两阶段LR-ADMM分解求解策略，先由交替方向乘子法获得算电协调初始点，再由拉格朗日松弛对带宽约束下的迁移方案进行细化。最后，以3个典型数据中心为算例进行验证。【结果】所提方法可使系统总成本降低14.10%，新能源消纳率由90.84%提升至97.09%，并在运行成本、服务质量和新能源消纳之间取得更优平衡。【结论】所提方法在所设计算例中获得了更低的总成本，并在服务质量代价与新能源消纳之间取得了更优平衡。
[Objective] With the continuous advancement of the national “Eastern Data， Western Computing” strategy， the spatiotemporal mismatch between the power consumption demand of multiple data centers （DCs） and renewable energy generation has become increasingly prominent. To address the inadequate characterization of data center power mapping and the imperfect coordination mechanism between the computing and power sectors in existing studies， this paper proposes a computing-power collaborative optimization method based on a two-stage Lagrangian relaxation-alternating direction method of multipliers （LR-ADMM） for spatiotemporal workload migration of multiple data centers. [Methods] Firstly， a refined power consumption model for data centers is established by considering the coupling characteristics of information technology （IT） equipment， auxiliary equipment and cooling systems. Meanwhile， constraints of service quality， network latency and link bandwidth are incorporated into the spatiotemporal workload migration model. Secondly， a collaborative optimization model is constructed with the objective of minimizing the total cost， including power purchase cost， carbon emission cost， wind and solar power curtailment cost， workload migration cost， and service quality loss cost. Thirdly， a two-stage LR-ADMM decomposition and solution strategy is designed. Specifically， the alternating direction method of multipliers （ADMM） is adopted to obtain the initial feasible solution for computing-power coordination， and Lagrangian relaxation （LR） is further applied to refine the workload migration scheme under link bandwidth constraints. Finally， three typical data centers are taken as numerical examples to verify the effectiveness of the proposed method. [Results] The proposed method reduces the total system cost by 14.10% and increases the renewable energy accommodation rate from 90.84% to 97.09%， which achieves a superior trade-off among operational cost， service quality and renewable energy utilization. [Conclusions] The proposed method realizes lower total system cost in the numerical cases and achieves an optimal balance between service quality loss and renewable energy accommodation capability.
数据中心（DC） / 工作负载时空迁移 / 算电协同 / 拉格朗日松弛-交替方向乘子法（LR-ADMM） / 新能源消纳
data center （DC） / spatiotemporal workload migration / computing-power coordination / Lagrangian relaxation-alternating direction method of multipliers （LR-ADMM） / renewable energy accommodation
| [1] |
International Energy Agency. Energy and AI[R]. Paris: IEA, 2025.
|
| [2] |
国家发展改革委等部门关于深入实施“东数西算”工程加快构建全国一体化算力网的实施意见[EB/OL]. (2023-12-25) [2026-06-05]. https://www.gov.cn/zhengce/zhengceku/202401/content_6924596.htm.
|
| [3] |
国家数据局. 数字中国发展报告(2024年)[R]. 北京: 国家数据局, 2025.
|
| [4] |
中国信息通信研究院. 算力电力协同发展研究报告(2025年)[R]. 北京: 中国信息通信研究院, 2025.
|
| [5] |
国家发展改革委, 工业和信息化部, 国家能源局, 等. 关于印发《数据中心绿色低碳发展专项行动计划》的通知[EB/OL]. (2024-07-03) [2026-06-05]. https://www.gov.cn/zhengce/zhengceku/202407/content_6964165.htm
|
| [6] |
陈敏, 高赐威, 郭庆来, 等. 互联网数据中心负荷时空可转移特性建模与协同优化: 驱动力与研究架构[J]. 中国电机工程学报, 2022, 42(19): 6945-6957.
|
| [7] |
丁肇豪, 曹雨洁, 张素芳, 等. 能源互联网背景下数据中心与电力系统协同优化(一): 数据中心能耗模型[J]. 中国电机工程学报, 2022, 42(9): 3161-3176.
|
| [8] |
杨挺, 姜含, 侯昱丞, 等. 基于计算负荷时-空双维迁移的互联多数据中心碳中和调控方法研究[J]. 中国电机工程学报, 2022, 42(1): 164-176.
|
| [9] |
|
| [10] |
|
| [11] |
|
| [12] |
|
| [13] |
|
| [14] |
|
| [15] |
向世林, 向月, 王晏亮, 等. 计及负载响应特性的多数据中心时空协同运行优化策略[J]. 中国电力, 2025, 58(4): 170-181.
|
| [16] |
刘方, 陈祥洲, 周江昕, 等. 考虑数据中心共享储能与计算负荷时空迁移特性的虚拟电厂优化运行方法[J]. 电网技术, 2026, 50(5): 1848-1864.
|
| [17] |
周钱雨凡, 杨苹, 万思洋, 等. 算力电力节点可调节资源的双向协同优化调度[J]. 电力建设, 2025, 46(2): 13-25.
|
| [18] |
周世博, 周明, 孙黎滢, 等. 激发多元灵活性的数据中心协同优化运行方法[J]. 电网技术, 2024, 48(11): 4417-4426.
|
| [19] |
唐伟, 单葆国, 郑海峰, 等. 数据中心两阶段源荷协同优化调度研究[J]. 电力系统及其自动化学报, 2023, 35(4): 49-58.
|
| [20] |
王继东, 许秋铭, 黄婷, 等. 含共享储能的数据中心微网群分布式优化调度[J]. 电网技术, 2024, 48(8): 3238-3247.
|
| [21] |
王丹阳, 张沈习, 程浩忠, 等. 考虑数据中心用能时空可调的多区域能源站协同规划[J]. 电力系统自动化, 2023, 47(3): 77-85.
|
| [22] |
|
| [23] |
张晓烽, 傅昂, 刘玉婷, 等. 考虑自然冷源的数据中心综合能源系统双层协同优化配置研究[J]. 太阳能学报, 2025, 46(9): 696-705.
|
| [24] |
吴成邦, 程志江, 陆海峰, 等. 考虑隐私保护的含互联网数据中心的综合能源系统时空联合规划[J]. 电力建设, 2025, 46(4): 84-98.
|
| [25] |
|
| [26] |
|
| [27] |
|
| [28] |
马浩然, 袁至, 王维庆, 等. 考虑数据中心和储能接入的主动配电网经济调度研究[J]. 发电技术, 2025, 46(4): 748-757, 634.
目的 为了提高电网新能源消纳能力、降低发电企业运营成本，解决大规模可再生能源馈入配电网时网损增大和新能源消纳率降低等问题，提出一种考虑数据中心(data center，DC)和储能协同调度的主动式配电网经济调度策略。 方法 首先，通过利用DC的时空可调特性，以及储能设备能够缓解电网发用电两侧矛盾的作用，充分协同调度DC和储能来达到削峰填谷；其次，为提高算法的搜索性能，在麻雀搜索算法中引入了Tent混沌初始化和多种群竞争等机制；然后，考虑到需求响应机制分时电价的影响，在储能成本、运行成本等多个约束条件下利用改进型麻雀搜索算法对模型进行优化调度，最大程度实现灵活性负荷与新能源的最优化匹配；最后，以改进IEEE 30和33节点系统为算例，仿真验证所提策略的有效性。 结果 提出的协同调度策略能够有效降低配电网综合运行成本，提高新能源消纳率。 结论 该策略可以为配电网经济调度的相关研究提供一定参考价值。
Objectives To improve the new energy consumption capacity of the power grid, reduce operational costs for power generation companies, and address issues such as increased network losses and reduced new energy consumption when large-scale renewable energy is integrated into the distribution networks, an economic scheduling strategy for active distribution networks with the inclusion of the coordinated scheduling of data center (DC) and energy storage is proposed. Methods Firstly, by utilizing the spatiotemporal adjustability of DC and the role of energy storage devices in alleviating the imbalance between grid supply and demand, the coordinated scheduling of DC and energy storage is fully leveraged to achieve peak shaving and valley filling. Secondly, the search performance of the sparrow search algorithm is improved by incorporating Tent Chaos initialization and multi-population competition mechanisms into the algorithm. Then, considering the impact of time-of-use pricing on the demand response mechanism, the model is optimized using an improved sparrow search algorithm under constraints such as energy storage costs and operational costs, aiming to achieve the optimal matching of flexible loads and new energy. Finally, the improved IEEE 30 and IEEE 33 node systems are taken as examples to verify the effectiveness of the proposed strategy. Results The proposed cooperative scheduling strategy can effectively reduce the overall operational costs of the distribution networks and improve the new energy consumption rate. Conclusions The strategy can provide some reference value for the related research of economic dispatch of distribution network. |
| [29] |
帅文利, 杨秀, 赵晓莉, 等. 计及数据中心负载特性的多时间尺度优化调度[J]. 智慧电力, 2025, 53(7): 1-10.
|
| [30] |
刘宗扬, 陈盛燃, 郑俊健, 等. 基于多站融合的新能源电站协调电压控制研究[J]. 电网与清洁能源, 2025, 41(6): 124-130.
|
| [31] |
范宏, 徐涛, 贾庆山. 数据-模型混合驱动的数据中心综合能源系统优化调度综述[J]. 南方电网技术, 2025, 19(3): 174-187.
为了实现数据中心的节能和减排，将数据中心融于综合能源系统进行协同优化是实现这一目标的有效途径。首先阐述了数据-模型混合驱动策略的原理，并分别对模型驱动和数据驱动在综合能源系统中的应用情况进行综述。然后，详细介绍了数据中心的负荷预测模型以及数据-模型混合驱动在数据中心综合能源系统中的应用现状，在此基础上，提出了一种基于数据-模型混合驱动的数据中心综合能源系统调度策略框架。最后，对目前研究存在的问题进行了讨论，并对未来的发展方向做出展望，为该领域的研究人员提供参考。
To achieve energy efficiency and emissions reduction in data centers, integrating data centers into integrated energy systems for collaborative optimization is an effective approach to achieve this goal. Firstly, the principles of a hybrid data-model driven strategy are explained, then the applications of model-driven and data-driven approaches in integrated energy systems is reviewed separately. Secondly, the load forecasting model of the data center and the current status of the application of a hybrid data-model driven method in integrated energy systems with data centers are introduced in detail. On this Basis, a framework for scheduling strategies in integrated energy systems with data centers utilizing a data-model hybrid drive is proposed. Finally, the current research issues are discussed, and an outlook on future development directions is provided, providing a reference for researchers in this field. |
| [32] |
杨秀, 张相寅, 黄海涛, 等. 基于多智能体近端策略网络的数据中心双层优化调度[J]. 南方电网技术, 2025, 19(4): 107-121, 131.
随着新一代信息通信技术，如5G、云计算和人工智能的不断演进，世界正迅速迈入数字经济的快车道。针对数据中心中可再生能源和工作负载预测的不确定性，提出了一种基于多智能体近端策略网络的数据中心双层优化调度方法。首先，建立了数据中心双层时空优化调度框架，对数据中心工作负载、IT设备、空调设备进行详细建模；在此基础上，提出数据中心的双层优化调度模型，上层以互联网数据中心（Internet data center，IDC）运营管理商总运营成本最小为目标进行时间维度调度，下层以各IDC运行成本最低为目标进行空间维度调度；然后，介绍多智能体近端策略网络算法原理，设计数据中心双层优化调度模型的状态空间、动作空间和奖励函数。最后，针对算例进行离线训练和在线调度决策，仿真结果表明，所提模型和方法能够有效降低系统成本和能耗，实现工作负载的最佳分配，具有较好的经济性和鲁棒性。
With the continuous evolution of new generation information and communication technologies such as 5G, cloud computing, and artificial intelligence, the world is rapidly entering the fast lane of the digital economy. A dual-layer optimization scheduling method for data centers based on multi-agent proximal strategy network is proposed to address the uncertainty of renewable energy and workload prediction in data centers. Firstly, a dual-layer spatiotemporal optimization scheduling framework for data centers is established, which provides detailed modeling of data center workloads, IT equipment, and air conditioning equipment; On this basis, a dual-layer optimization scheduling model for data centers is proposed. The upper layer schedules the time dimension with the goal of minimizing the total operating cost of IDC operators, while the lower layer schedules the space dimension with the goal of minimizing the operating cost of each IDC. Then, the principle of multi-agent proximal strategy network algorithm is introduced, and the state space, action space, and reward function of the dual-layer optimization scheduling model for data centers are designed. Finally, offline training and online scheduling decisions are conducted for the examples. Simulation results show that the proposed model and method can effectively reduce system costs and energy consumption, achieve optimal workload allocation, and have good economy and robustness. |
| [33] |
韩雪姣, 屈鲁, 王长永, 等. 数据中心全直流供电系统的构建及其综合评价[J]. 浙江电力, 2024, 43(3): 75-83.
|
| [34] |
杨学佳, 孙浩, 钱健, 等. 基于改进NSGA-II算法的数据中心能量调度优化研究[J]. 电力需求侧管理, 2025, 27(4): 64-70.
|
| [35] |
文淅宇, 朱继忠, 李盛林, 等. 基于时空协同的多数据中心虚拟电厂低碳经济调度策略[J]. 电力系统自动化, 2024, 48(18): 56-65.
|
| [36] |
张文翰, 周海浪, 尤培培, 等. 考虑带宽和电力费用的数据中心工作流负载分配优化模型研究[J]. 电力需求侧管理, 2024, 26(3): 48-54.
|
| [37] |
朱子恒, 张策, 丁肇豪, 等. 数据中心纳入全国碳排放权交易市场机制研究[J]. 中国电机工程学报, 2024, 44(14): 5562-5573.
|
| [38] |
丁巧宜, 王梓耀, 潘振宁, 等. 面向电量-调频-容量市场的数据中心园区算力及电力资源规划[J]. 电力系统自动化, 2024, 48(1): 59-66.
|
| [39] |
罗清局, 朱继忠. 基于改进交替方向乘子法的电-气综合能源系统优化调度[J]. 电工技术学报, 2024, 39(9): 2797-2809.
|
| [40] |
|
| [41] |
工业和信息化部国家机关事务管理局国家能源局关于加强绿色数据中心建设的指导意见[EB/OL]. (2019-01-21)[2026-06-05]. https://www.gov.cn/zhengce/zhengceku/2019-10/21/content_5442976.htm.
|
所有作者声明不存在利益冲突。
孙亮提出研究方向，设计论文框架；杨世龙设计研究思路、研究方案，对研究方案可行性进行调查分析，实施研究过程；党翠收集数据，采集、清洗与分析数据；张宗振参与论文审核；曹鑫雨参与论文最终版本的修订；江守其参与论文最终版本的修订。所有作者均阅读并同意了论文终稿内容。
AI小编
/
| 〈 |
|
〉 |
AI 导读
