Simscape steady state modeling and analysis on the dynamic behavior of a cooling distribution unit in a data center

The rapid growth of high-performance computing and artificial intelligence (AI) workloads in modern data centers has led to increased heat densities, posing significant challenges for maintaining reliable and efficient cooling. Single-phase indirect liquid cooling, supported by Cooling Distribution Units (CDUs), offers a promising solution due to its high thermal efficiency and compact design. However, understanding the transient thermal behavior of such systems under variable heat loads remains limited, as most studies focus primarily on steady-state performance. This thesis presents a MatLab Simscape model of a single-phase indirect liquid cooling CDU, developed using manufacturer data and validated against experimental measurements. Steady-state simulations demonstrate excellent agreement with measured temperatures, volumetric flow rates, pressure drops, and heat rates, with error percentages approximately negligible. Subsequently, dynamic performance is evaluated through heat rate step-load experiment and by applying realistic AI heat rate step-load profiles. The transient responses predicted by the model closely match experimental results, providing insight into the system’s thermal inertia and stability under rapidly changing conditions. The validated model enables a detailed analysis of the CDU’s transient thermal response, highlighting the system’s behavior during sudden variations in heat loads. The findings contribute to improved understanding of liquid cooling system dynamics and provide a foundation for future optimization of CDU design and operation in high-performance data centers.