Dynamic Modeling of AI Data Center Loads and Grid Stability Studies

Xin Chen – TAMU

AI data centers are emerging as large, power-electronic–dominated loads whose fast disturbance-time behavior is not well captured by conventional static or aggregate load models, creating uncertainty in interconnection and stability studies. This seminar presents our ongoing work on high-fidelity electromagnetic transient (EMT) dynamic modeling of AI data center loads for grid studies. We develop detailed EMT representations of the data center electrical hierarchy, including the point of interconnection, transformers, UPS systems, cooling systems, power converters, computing server loads, and associated protection and control systems. The models capture grid-relevant behaviors such as rapid active and reactive power response, voltage- and frequency-dependent characteristics, ride-through performance and protection logic, and operating-mode transitions. These capabilities enable improved understanding of AI data center dynamics and their implications for grid stability.