电力系统自动化2026，Vol.50Issue(17)：85-95,11.DOI:10.7500/AEPS20251113003
计及计算资源自适应调整的数据中心算力-电力协同调度
Coordinated Scheduling of Computing Power and Electric Power in Data Centers Considering Adaptive Adjustment of Computing Resources
摘要
Abstract
Currently,leveraging the transferable features of computation tasks in data centers to enhance quality of service and reduce operating costs has become a significant development direction in data center operations.However,as major energy consumers,the operation modes of data centers significantly impact the safe operation of distribution networks.Therefore,focusing on the coordinated scheduling of data center clusters and distribution networks,the computation task transfer is integrated with power system operation to construct a hierarchical reinforcement learning model.The model adopts an adaptive adjustment mechanism for computing resources and establishes a hierarchical coordinated framework that incorporates economic dispatch on the electric-power side and task transfer on the computing-power side.On this basis,a hierarchical multi-agent reinforcement learning algorithm based on abstract instructions is designed to generate power exchange instructions between the distribution network and data centers,task transfer commands,and computing resource scheduling strategies.Simulation results demonstrate that the proposed method can effectively reduce system economic costs,enhance renewable energy accommodation capability,significantly decrease execution violations of computation tasks,and improve overall quality of service.关键词
算力-电力协同/调度/强化学习/数据中心/计算资源/配电网/任务迁移/服务质量Key words
computing power-electric power coordination/scheduling/reinforcement learning/data center/computing resource/distribution network/task transfer/quality of service引用本文复制引用
张洋洋,涂正宏,徐英,仪忠凯,林晨浩..计及计算资源自适应调整的数据中心算力-电力协同调度[J].电力系统自动化,2026,50(17):85-95,11.基金项目
国家自然科学基金资助项目(U25A20337).This work is supported by National Natural Science Foundation of China(No.U25A20337). （U25A20337）
