面向算电协同的智算中心负荷灵活性:挑战与展望OA

The prevalence of artificial intelligence(AI)has led to a sharp increase in energy consumption and carbon emissions in intelligent computing centers.High volatility on both the resource and load sides makes the exploration of endogenous load flexibility the key to computing power-electric power coordination.First,addressing the unique characteristics of AI computing tasks and hardware infrastructure,this paper analyzes the three major challenges faced by the research on load flexibility of intelligent computing centers:multi-dimensional heterogeneity,fine-grained coupling,and multi-stakeholder collaboration.Then,based on the three core sub-problems of load modeling,flexibility exploitation,and flexibility utilization,this paper reviews the current status of domestic and international research.Finally,the paper prospects future research trends,highlighting the need to advance task-level load modeling adapted to multi-dimensional heterogeneity,conquer dynamic flexibility characterization technologies that perceive fine-grained coupling,and design flexibility expansion application strategies to incentivize multi-stakeholder collaboration.