面向大规模边缘数据中心的分布式均衡协同调控策略

A Distributed Equilibrium Collaborative Scheduling Strategy for Large-Scale Edge Data Centers

中文关键词:边缘数据中心  需求响应  聚合博弈  纳什均衡  动态电价

keywords:edge data center  demand response  cooperative game  Nash equilibrium  dynamic price

中文摘要:针对现有数据中心调度策略存在物理建模粗糙、系统可扩展性差，且易受分时电价影响引发“峰谷倒置”的多重局限，提出一种计及精细化物理特性的边缘数据中心分布式协同调度方法。首先，通过精细化建模批处理任务特性与热惯性，明确系统灵活性边界；其次，构建基于动态电价的多边缘数据中心非合作博弈模型，设计分布式迭代算法求解纳什均衡，有效规避集中式调度的计算压力与数据隐私泄露风险。最后，基于300个数据中心的仿真实验结果表明，所提方法可有效平抑负荷波动、充分释放系统调节潜力，同时能彻底避免因响应分时电价而产生的二次负荷尖峰，提升调度的经济性与稳定性。

Abstract:Existing scheduling strategies for data centers suffer from rough physical modeling，poor system scalability，and peak-valley inversion easily caused by time-of-use electricity prices. To address the above limitations，this paper proposes a distributed collaborative scheduling method for edge data centers considering refined physical characteristics. First，refined modeling is performed for the characteristics of batch processing tasks and system thermal inertia to define the flexibility boundary of the data center system. Second，a non-cooperative game model for multiple edge data centers is established based on dynamic electricity prices，and a distributed iterative algorithm is designed to solve the Nash equilibrium. This method effectively relieves the computational pressure of centralized scheduling and avoids the risk of data privacy leakage. Finally，simulation experiments conducted on 300 data centers verify that the proposed method can effectively smooth system load fluctuations and fully exploit the regulatory potential of the system. Furthermore，it thoroughly eliminates secondary load spikes triggered by responses to time-of-use electricity prices，thereby improving the economic efficiency and operational stability of system scheduling.