Research on cascade energy utilization in artificial intelligence data center energy systems for regional energy supply

Artificial intelligence data centers face high energy consumption and carbon emissions. Developing energy systems that couple cascade energy utilization with waste heat recovery has thus become a key research direction. However, existing studies commonly exhibit imprecise thermal-grade matching and neglect heat-transfer-medium mass balance under off-design conditions, while conventional scheduling methods are inadequate for handling the inherent nonlinear operating characteristics of such systems. To this end, this study designs an artificial intelligence data center energy system that integrates a multi-stage combined cycle, high-temperature heat pumps, photovoltaics, and hybrid energy storage. Based on physical mechanisms, thermodynamic models of key equipment are established, providing computational support for the formulation of regional multi-load energy supply schemes. Meanwhile, a hybrid scheduling strategy combining heuristic decision-making with mathematical optimization approaches is proposed to coordinate flexible resources such as energy storage and the power grid. A case study of the proposed system in Beijing shows that the regional multi-grade energy supply scheme reduces carbon emissions by over 7%. Cascade energy utilization increases system revenue by 16.35% and exergy efficiency by 3.53%. Waste-heat upgrading by high-temperature heat pumps reduces the regional heating deficit by more than 60%. In addition, the proposed hybrid scheduling strategy fully exploits the flexible regulation potential of energy storage and reduces the typical daily operating cost by 4.91%–7.23%. Sensitivity analyses across diverse climate zones further validate the robustness of the findings.