面向算电协同的建模与应用:研究现状与展望OA

The rapid proliferation of artificial intelligence has significantly increased global computing power demands,drawing widespread attention to the coordination of data centers and power systems.To address this,this paper conducts a systematic study on computing power-electric power coordination.First,it summarizes fundamental models of data centers in computing resource allocation,energy consumption management,and thermal management,and synthesizes modeling that characterizes the coupling between data centers and various energy networks,which lays a foundation for the research on computing power-electric power coordination.It then focuses on research advances in long-term expansion planning and short-term optimal operation for computing power-electric power coordination in three representative scenarios,such as data center microgrids.Furthermore,the paper provides a comprehensive overview of relevant market mechanisms from five perspectives,including the electricity and flexibility markets.It also summarizes resilience enhancement measures,mathematical models,and solution algorithms for computing power-electric power coordination at both transmission and distribution network levels.Subsequently,to further promote the research in this area,the paper surveys and compiles key characteristics of mainstream open data center datasets,including temporal granularity,coverage period,features,and applicable scenarios.Finally,it outlines key research directions for future computing power-electric power coordination from both micro-level aspects,such as coordinated hardware deployment in AI data centers and AI-driven load demand response,and macro-level aspects,such as social and economic benefits.