电力-算力协同:内涵、架构、技术与实践OA

The rapid evolution of artificial intelligence technology has driven a sharp rise in the load on computing infrastructure,represented by data centers.Currently,the expansion of green power supply and computing scale in China faces challenges of spatial resource mismatch and temporal supply-demand misalignment.Consequently,electricity-computing power synergy has emerged not only as the key to supporting green computing with green power but also as an inevitable path for the deep integration of China's"East Data,West Computing"and"West-East Power Transmission"national projects.This paper first elucidates the connotation and evolutionary path of electricity-computing power synergy,identifying two key scientific problems:mapping and representation of heterogeneous electricity-computing resources,and joint minimization of energy entropy and information entropy.Subsequently,a six-layer technical architecture spanning entity,sensing,simulation,decision,execution,and incentive layers is constructed.Based on this framework,the paper proposes core research directions for key technologies,including mapping representation and electrical equivalent modeling of computing-electricity resources,establishment of coupled digital twins and simulation deduction,collaborative resource regulation and online optimization-based decision-making,grid integration and reliable spatiotemporal migration of heterogeneous computing power,as well as incentive-compatible mechanisms and trusted trading in multi-dimensional markets.Finally,the paper systematically reviews current research status and practices in electricity-computing power synergy across multiple spatial scales,ranging from efficient energy use within data centers to park-level multi-energy complementarity,regional supply-demand coordination,and wide-area joint dispatching,and provides an outlook on the future development trends of key technologies.