Bi-Level Stackelberg Game-Based Optimization Model for Shared Energy Storage in Data Centers Considering Computational Flexibility

Abstract

With the explosive growth in demand for artificial intelligence and computing power, the energy consumption of data centers has sharply increased, making their green and low-carbon operation an urgent need. Shared energy storage (SES), as a flexible regulation resource, can effectively facilitate renewable energy consumption and reduce system costs. However, existing research mostly regards data center loads as rigid loads, ignoring the elastic scheduling potential of latency-insensitive computing tasks, and lacking a game decision-making model from the perspective of SES operators to provide strategies for data center SES transactions. Therefore, this study constructs a SES trading optimization model that takes into account the flexibility of computing power. The upper-level targets profit maximization for the SES operator by optimizing charge and discharge strategies and service pricing, while the lower level minimizes the total energy cost of each computing center by jointly optimizing computing task scheduling and energy storage utilization plans. Finally, a bi-level Stackelberg game-based optimization model is constructed to solve the SES dispatch and, achieve benefit coordination between SES and computing center. The simulation results indicate that using SES without implementing load shifting cannot optimize the total cost of the computing power center, offering only storage revenue. Only by integrating SES with load shifting can a significant cost reduction be realized. The model presented in this article resulted in an SES revenue of 2605.34 yuan and a 3.14% reduction in computing center costs. This study provides a theoretical basis and decision-making support for shared energy storage participation in the computing power market and contributes to accelerating the coordinated development of computing power and electricity systems.