Co-allocation and operational optimization for green power-direct-supply data center clusters with shared energy storage

• Two-stage, bi-level collaborative framework for optimal data center cluster (DCC) configuration and operation. • Data-driven sample-based robust optimization (DSRO) addresses renewable generation and workload uncertainties. • Asymmetric Nash bargaining with a novel logarithmic function ensures fair benefit allocation among DCC. The rapid proliferation of data center cluster (DCC) presents significant challenges to power system flexibility and renewable energy integration. This paper proposes a two-stage, bi-level collaborative framework to optimize the joint operation and capacity planning of DCC in Green Power Direct Supply (GPDS) mode and a Shared Energy Storage system. To manage uncertainties in renewable generation and workloads, a data-driven sample-based robust (DSRO) optimization approach is employed to mitigate the conservatism of traditional methods. In the first stage, a robust configuration model determines the optimal capacities of the SES and gas turbines, while linearizing the M/M/1 queuing model to guarantee a service violation rate below 5%. The second stage focuses on distributed operation. The complex, non-convex problem is decoupled into cost minimization and benefit allocation subproblems, solved efficiently using the alternating direction method of multipliers. Cooperative benefits are then distributed via an asymmetric Nash bargaining game, where a novel logarithmic function quantifies each participant’s bargaining power based on their net energy contribution, ensuring fair allocation. Simulation results validate that the proposed framework significantly enhances system economy, renewable accommodation, and resource allocation flexibility, offering a robust foundation for multi-agent collaborative energy systems.