Game-Based Optimization Method for Geo-Distributed Data Centers under Customer Directrix Load Demand Response Mechanism

This study considers geo-distributed data centers (DCs) with diverse load profiles, electricity prices, and renewable energy availability, aggregated within their respective local virtual power plants (VPPs). To optimize demand response (DR) within such a system, a two-stage decision-making framework is proposed. First, a potential game-based method is developed to decompose the total customer directrix load (CDL) into regional sub-CDLs, aligning with grid requirements while minimizing VPP adjustment costs. Second, an evolutionary game model optimizes workload allocation strategies for DCs across spatial and temporal dimensions, considering constraints of bounded rationality and incomplete information. The proposed models are rigorously analyzed for equilibrium existence, convergence, and stability. A case study based on Google’s geo-distributed DCs and local grid data demonstrates that the framework effectively enhances VPP revenues and reduces DC operational costs, especially during peak demand periods.