Multi-Source Energy Storage Load-Based Task Scheduling for Data Centers

As data center energy consumption escalates, the associated operational costs and impact on the power grid have become critical issues. To address these challenges, this paper proposes a coordinated optimization framework that integrates multi-source energy storage with computing task scheduling. First, a data center microgrid model comprising Photovoltaic (PV) generation, Battery Energy Storage Systems (BESS), and water-based Thermal Energy Storage (TES) is established. Specifically, a probabilistic mixture model based on meteorological data from Beijing is employed to characterize and mitigate the uncertainty of PV output. Second, computing workloads are categorized into three classes - urgent, semi-urgent, and flexible - utilizing load shifting matrices to fully exploit the potential of demand-side response. Subsequently, a scheduling model based on linear programming is developed with the objective of minimizing daily operational costs. Case studies demonstrate that this strategy effectively leverages Time-of-Use (ToU) pricing to achieve "peak shaving and valley filling". Simulation results indicate that, compared to single-optimization methods, the proposed coordinated strategy reduces daily operating costs by approximately 7.5%, validating the model's effectiveness in enhancing both the economic viability of data centers and their grid friendliness.