电力系统自动化2026，Vol.50Issue(17)：62-73,12.DOI:10.7500/AEPS20251126001
合作博弈视角下计算任务调度驱动的算力-电力双向协同优化
Computing-task-scheduling-driven Bidirectional Coordinated Optimization for Computing Power and Electric Power from Perspective of Cooperative Game
摘要
Abstract
Data centers can adjust their load through computing task scheduling,which will become an important flexible demand-side regulation resource in new power systems.However,data centers and power systems have different operation objectives and cannot fully share information.An effective bidirectional coordination method for computing power and electric power has not yet been established.This paper proposes a bidirectional coordinated optimization method for computing power and electric power driven by computing task scheduling from a cooperative game perspective.First,a bidirectional coordinated optimization framework for data center operators and power system operators is proposed based on the Nash bargaining theory,to coordinate utility improvements for both parties.Next,detailed computing task scheduling models for interactive workloads and batch jobs are established,and a computing task scheduling decision-making method based on rolling horizon optimization is proposed.Subsequently,a distributed online solution method for collaborative decision-making of computing power and electric power is proposed.It balances global optimality and real-time solution efficiency while preserving the confidentiality of internal operation parameters and information of both parties.Case study results verify the effectiveness of the proposed method.关键词
数据中心/算力-电力协同/需求侧响应/合作博弈/滚动时域优化/计算任务调度Key words
data center/computing power-electric power coordination/demand-side response/cooperative game/rolling horizon optimization/computing task scheduling引用本文复制引用
王鹏,刘文宇,曹雨洁,孙逸萌,卫璇,丁肇豪..合作博弈视角下计算任务调度驱动的算力-电力双向协同优化[J].电力系统自动化,2026,50(17):62-73,12.基金项目
智能电网国家科技重大专项资助项目(2025ZD0804700).This work is supported by Smart Grid-National Science and Technology Major Project(No.2025ZD0804700). （2025ZD0804700）
