Day-ahead operation strategy for multiple data centre prosumers: A cooperative game approach

The transferrable characteristics of workloads both spatial and temporal have received much attention recently. A data centre prosumer (DCP) is created by combining multiple data centre buildings, which contain workloads, photovoltaic generations, and battery energy storage systems (BESSs). Few studies have considered electricity trading and payoff sharing among various DCPs in a data centre park. This paper presents a cooperative framework for interconnected DCPs in a data centre park and proposes a novel day-ahead operation strategy considering uncertainties in photovoltaic (PV) generations and workloads using the equivalent linear robust reformation method. The operation strategy is formulated as an asymmetric Nash bargaining cooperative game, which aims at maximizing the overall payoff of the cooperative alliance while considering the contributions of shared electricity to share the payoff. To solve the formulated cooperative game problem while preserving user privacy, an enhanced self-adaptive prediction-correction-based alternating direction multiplier method (PCB-ADMM) algorithm with adaptive penalty coefficients to accelerate convergence is employed in a distributed manner. The case studies indicate a 6.63% reduction in overall cost and verify the feasibility and fairness of the proposed cooperative strategy. The enhanced self-adaptive PCB-ADMM algorithm also shows superiority over other ADMM algorithms in effectiveness and rationality.