Digital Twin-Based Cooling System Optimization for Data Center

Data center cooling systems consume significant auxiliary energy, yet optimization studies rarely quantify the gap between theoretically optimal and operationally deployable control strategies. This paper develops a digital twin of the liquid cooling infrastructure at the Frontier exascale supercomputer, where a hot-water system comprises three parallel subloops, each serving coolant distribution unit clusters through plate heat exchangers and variable-speed pumps. The surrogate model is implemented in Modelica and validated against one year of 10-minute operational data following ASHRAE Guideline 14, achieving a coefficient of variation of the root mean square error below 2.7% and a normalized mean bias error within 2.5% at the subloop level. Using this validated model, a layered optimization framework evaluates three progressively constrained strategies: an analytical flow-only optimization yields 20.4% total energy savings, unconstrained joint optimization of flow rate and supply temperature achieves 30.1% savings, and ramp-constrained co-optimization that enforces actuator rate limits still delivers 27.8% savings. The analysis reveals that the baseline operation uses 2.9 times the minimum thermally safe flow rate and shows that co-optimizing supply temperature with flow rate nearly doubles the energy savings achievable by flow reduction alone.