Spatiotemporal Thermodynamic Dynamic Load Modeling and Characteristic Analysis of Data Centers for Power System Applications
Data center expansion has driven a surge in energy consumption, and prediction deviations for high-density loads introduce risks to power grid planning and operation. Existing models exhibit significant limitations: they neglect temperature non-uniformity and the nonlinear impact of rack layout on airflow; rely on static power usage effectiveness (PUE) without accounting for the coupling between load fluctuations and cooling dynamic response; and linearize cooling systems, ignoring IT equipment–cooling system–thermal environment interactions. This paper proposes a spatiotemporal thermodynamic dynamic load model for data centers based on computational fluid dynamics (CFD). Spatially, a physical model incorporating cold aisle containment, server power, and layout is constructed; CFD simulations capture temperature non-uniformity and establish a collaborative IT power–temperature field–cooling power model. Temporally, IT load time-series fluctuations, thermal inertia response, and cooling power dynamics are coupled to quantify the impact of physical parameters on overall load at short time scales. Simulation results for a typical day show that traditional models, by neglecting these key parameters, may cause instantaneous load discrepancies of 12.56%–34.21%. Data center load characteristics are strongly coupled with internal thermal environment dynamics, and the proposed model provides a high-fidelity thermodynamic reference for data centers participating in grid interaction as flexible resources.