A Two-Stage Demand Response Stackelberg Game of Data Center Operators and the System Operator Based on Kriging Metamodel

Due to the spatially and temporally transferable workloads, data centers (DCs) become increasingly important in demand response (DR) implementations. By making full use of their owned DCs, data center operators (DCOs) are significant DR resource providers. To fully exploit the DR capability of the DCO on different time scales, we present a two-stage scheduling model for DCOs and the system operator (SO). In the DR scheduling, the SO formulates DR compensation prices first, and then each DCO decides its best-response power demands accordingly. Considering the profit-hunting property of the SO and DCOs, a two-stage DR Stackelberg game model is proposed. Furthermore, the existence and uniqueness of the Stackelberg equilibrium are proved. Finally, to protect the data privacy of DCOs, we design a Kriging-metamodel-based algorithm which avoids the DCO privacy exposure to the SO in the optimization process. Simulation results prove the accuracy and the calculation efficiency of the proposed Kriging-metamodel-based algorithm.
Note to Practitioners
—DCs play an increasingly important role in the power balance of grids in recent years. Existing studies focusing DCs’ DR participation generally consider the single-stage DR participation of DCs which ignores the role of DCOs in unified scheduling and management of their owned DCs. Considering the impacts of service request submitting time on the DR capabilities of DCs, this work proposes a novel two-stage DR scheduling model in the account of the unified scheduling role of DCs. Then, a Stackelberg game model is further proposed to characterize the profit-hunting property of the SO and DCOs. Finally, a privacy-protected algorithm is designed to address the optimization problem without explosions of DCO’s sensitive data. Simulation results show the effectiveness of the proposed model and algorithm in encouraging DCOs to adjust their energy consumption plans according to the SO’s request. Moreover, it is proven that the proposed algorithm can achieve global optimization with effective data privacy protection and low computation cost. Simulation results suggest that the proposed method can obtain a high-quality solution for a Stackelberg game model with privacy protection.