摘要
在计算需求爆发式增长、算力资源成本高、数据中心能耗高以及电网运行稳定受到影响的背景下,亟需探索算力与电力可调节资源的双向协同技术以降低算力的能耗成本并提高电网运行的稳定性和经济性。首先,构建算力电力节点可调节资源的双向协同调度架构,分别对算力节点和电力节点内的多元实时可调资源进行量化建模。然后,考虑计算工作任务与算力资源的匹配性、实时调节特性,通过对算力节点下计算工作任务的调度以及电力节点下的可调节负荷调度,提出双层两阶段算电协同优化调度模型。最后,通过算例验证了算电节点可调节资源双向协同优化可行且效果显著:在电力节点整体可调节资源量约2 300 MW的设定下,50 MW算力节点为电力节点运行降低的成本可占到4.71%;同时算力节点自身日内运行成本降低约0.70%。
The era of intelligence has driven computing power resources to become highly flexible and adjustable.They have also made the bidirectional collaborative optimization of computing and electricity into a new method of economic optimization in comprehensive energy systems.The explosive growth in computational demand has led to a shortage of computational resources.It also brings the challenges of high energy consumption and carbon emissions for data centers,where the annual electricity consumption can reach billions of kilowatt-hours.When the cost of computing resources is high and the stability of power grid operations is affected,there is an urgent need to explore bidirectional collaborative technologies between computing and power nodes with adjustable resources to reduce the energy cost and enhance the stability and economic efficiency of power grid operation.This study constructs a bidirectional collaborative scheduling architecture for adjustable resources in computing and power nodes,and quantitatively models the diverse adjustable resources within them.Considering the matching and real-time adjustment characteristics between computing tasks and resources,a dual-layer two-stage collaborative optimization scheduling model is proposed by scheduling computing tasks under the computing node and adjustable loads under the power node.Through numerical examples,it was verified that the bidirectional collaborative optimization of adjustable resources for computing and power nodes is feasible and effective.Under the setting of an overall adjustable resource of approximately 2300 MW for power nodes,the cost reduction provided by computing nodes of 50 MW can account for 4.71%of the power node operation,while reducing its daily operating costs by approximately 0.70%.
作者
周钱雨凡
杨苹
万思洋
崔嘉雁
李丰能
隗知初
ZHOU Qianyufan;YANG Ping;WAN Siyang;CUI Jiayan;LI Fengneng;WEI Zhichu(School of Electric Power,South China University of Technology,Guangzhou 510641,China)
出处
《电力建设》
北大核心
2025年第2期13-25,共13页
Electric Power Construction
基金
国家自然科学基金项目(51937005)
国家重点研发计划资助项目(2023YFB4203102)。
关键词
算电协同
可调节资源
双层优化
优化调度
computing and power collaboration
adjustable resources
double-layer optimization
optimize scheduling
