首页|期刊导航|综合智慧能源|考虑算力负载时空迁移特性的多数据中心低碳经济调度

考虑算力负载时空迁移特性的多数据中心低碳经济调度OA

Low-carbon economic scheduling for multiple data centers considering spatiotemporal migration characteristics of computing loads

中文摘要英文摘要

"东数西算"战略背景下,多数据中心算力负载的时空灵活性尚未得到充分利用,且固定碳配额难以匹配不同区域与典型日的碳排放特征,导致系统经济性与低碳性未能达到最优.面向多数据中心微电网系统,建立考虑区域气候、电价和可再生能源禀赋差异的数据中心能耗模型,将算力负载划分为交互式负载和批处理负载,其中交互式负载须在到达时段由本地即时消纳,批处理负载具备时空可调节特性.在此基础上,构建以源数据中心、到达时段、目标数据中心和执行时段为维度的四维算力迁移模型,在同一优化框架内联合确定批处理负载的执行时间与执行位置,从而反映区域电价、可再生能源出力与链路传输条件之间的耦合关系.以系统总成本最小为目标,建立计及购电、运维、碳交易和负载迁移成本的混合整数线性规划模型,采用YALMIP调用CPLEX求解.设置负载不可迁移、仅时间迁移和时空协同迁移3种递进式情景进行对比分析,以揭示算力迁移对系统经济性、碳排放和可再生能源消纳的增量影响.此外,在年度碳配额总量不变的前提下,提出基于调度前碳排放压力预测的典型日自适应碳配额分配方法,并与固定基准分配方案进行比较.算例结果表明:与负载不可迁移情景相比,时空协同迁移使系统总成本和碳排放量分别降低了9.41%和11.50%;负载迁移方向能够响应区域电价和可再生能源出力差异,日间负载主要向光伏资源充裕的数据中心转移,夜间负载主要向风电资源充裕的数据中心转移,光伏和风电消纳率分别达到99.84%和100%;在年度总配额不变条件下,典型日自适应分配使年度碳交易成本和系统总成本分别降低6.19%和0.25%.时空协同迁移能够有效匹配算力负载与低价低碳电力资源,实现算力与电力的协同优化;自适应配额分配能够改善配额与典型日排放压力的匹配关系,在不收紧年度碳排放约束的前提下降低碳交易履约成本,为地理分布式数据中心的低碳运行提供了实用化途径.

In the context of the"East Data,West Computing"strategy,the spatiotemporal flexibility of computing loads across multiple data centers has not been fully utilized,and fixed carbon quotas fail to match the carbon emission characteristics of different regions and typical days,resulting in suboptimal economic and low-carbon performance of the system.For a multi-data-center microgrid system,a data center energy consumption model considering differences in regional climate,electricity prices,and renewable energy endowments was developed.Computing loads were divided into interactive and batch loads.Interactive loads had to be processed locally and immediately during their arrival periods,while batch loads had spatiotemporal adjustability.On this basis,a four-dimensional computing load migration model was constructed with the source data center,arrival period,destination data center,and execution period as its dimensions.The execution time and location of batch loads were jointly determined within the same optimization framework,thereby reflecting the coupling relationships among regional electricity prices,renewable energy output,and link transmission conditions.With the objective of minimizing total system cost,a mixed-integer linear programming model considering electricity purchase,operation and maintenance,carbon trading,and load migration costs was established and solved using CPLEX called via YALMIP.Three progressive scenarios—no workload migration,temporal migration only,and coordinated spatiotemporal migration—were compared to reveal the incremental effects of computing load migration on system economic performance,carbon emissions,and renewable energy consumption.In addition,under the premise of keeping the total annual carbon quotas constant,an adaptive typical-day carbon quota allocation method based on pre-scheduling carbon emission pressure prediction was proposed and compared with a fixed baseline allocation scheme.Case study results showed that,compared with the no-load-migration scenario,coordinated spatiotemporal migration reduced total system cost and carbon emissions by 9.41%and 11.50%,respectively.The direction of load migration responded to differences in regional electricity prices and renewable energy output.Daytime loads were transferred mainly to data centers with abundant photovoltaic resources,whereas nighttime loads were transferred mainly to those with abundant wind resources.The photovoltaic and wind power consumption rates reached 99.84%and 100%,respectively.Under the condition that the annual total carbon quota remained unchanged,the typical-day adaptive allocation reduced annual carbon trading cost and total system cost by 6.19%and 0.25%,respectively.Coordinated spatiotemporal migration can effectively match computing loads with low-cost,low-carbon electricity resources,achieving coordinated optimization of computing power and electricity.The adaptive quota allocation can improve the matching relationship between quotas and typical-day emission pressure.It reduces carbon trading compliance costs without tightening annual carbon emission constraints,thereby providing a practical pathway for low-carbon operation of geographically distributed data centers.

付利卓;王成福;鲁浩;蔡继增

山东大学 电气工程学院,济南 250061山东大学 电气工程学院,济南 250061山东电力工程咨询院有限公司,济南 250013山东大学 电气工程学院,济南 250061

能源科技

多数据中心算电协同算力负载时空迁移低碳经济调度碳交易碳配额分配可再生能源消纳

multiple data centerscomputing-electricity coordinationspatiotemporal migration of computing workloadslow-carbon economic schedulingcarbon tradingcarbon quota allocationrenewable energy consumption

《综合智慧能源》 2026 (9)

84-95,12

国家自然科学基金项目(52377108) National Natural Science Foundation of China(52377108)

10.3969/j.issn.2097-0706.2026.09.007

评论