Spatiotemporal Thermodynamic Dynamic Load Modeling and Characteristic Analysis of Data Centers for Power System Applications

• • Traditional models (PUE, RC) that ignore server layout, cold aisle containment, and load time-series cause instantaneous load discrepancies of 12.56%–34.21%, directly undermining peak load and ramp-rate analyses for grid planning and demand response. • • The CFD-based model reveals a nonlinear dynamic coupling among IT load, temperature field, and cooling power, demonstrating that data center load is dominated by internal thermophysical dynamics rather than a linear mapping of instantaneous IT load. • • Compared to static PUE models, the proposed model captures thermal inertia-induced power delays and dynamic fluctuations; unlike second-order RC gray-box models, it quantifies the decisive influence of physical parameters such as server layout on non-uniform temperature fields and local hot spots. • • The model enables deep coupling with grid planning models via constraints, boundaries, and variable transfer, supporting second-level (or even millisecond-level) transient simulations integrated into long-term planning—a recent hotspot in equipment planning research.