算力-电力耦合网络的统一图模型及协同优化机制OA

Computing power-electric power coupling network faces challenges such as model heterogeneity,computational complexity,and market incentive mechanisms.To address these issues,a graph modeling method for the computing power network is first proposed to express the computing power constraints and the mapping relationship of computing power-electric power regulation capabilities.Based on this,a unified graph model representation for both computing power and electric power networks is proposed,from which a scalable distributed coordinated optimization method and analytical solutions for non-energy-storage devices are derived.This method supports network splitting at the computing power-electric power interface to achieve privacy-preserving computation.Based on this coordinated optimization capability and the Nash bargaining method,an incentive mechanism for the spatial migration of computing power is proposed,which functions independently of nodal electricity prices.Simulation results demonstrate that the proposed distributed optimization method exhibits linear complexity.Furthermore,the proposed computing power-electric power coordinated incentive mechanism can fairly allocate the cooperative surplus generated by spatio-temporal migration,achieving a win-win outcome for both the power grid and data centers.It also effectively enhances the renewable energy accommodation rate while reducing system operation costs and carbon emissions.