Collaborative Research: Unified Modeling of Grid-Data Center Power Electronics Interactions for Reliable and Stable Power Systems

Artificial intelligence, cloud computing, and large-scale data centers are rapidly increasing electricity demand and transforming the way power systems operate. Data centers are large electric loads, but the electric grid does not directly see the raw power demand created by computing tasks. Instead, that demand is filtered and reshaped by the power converters, control loops, and energy-storage elements within the data center before it reaches the grid connection point. Existing planning and operational tools often treat data centers using static or simplified models, failing to capture rapid, workload-driven load swings and their impact on grid stability. This project aims to establish a new framework for understanding how internal power-electronic systems transform computing activity into the effective load and impedance seen by the electric grid. The project will bring transformative change by revealing this connection and helping improve the reliability and stability of future energy infrastructure that supports artificial intelligence and advanced computing. The broader impacts of the project include advancing the scientific foundation for reliable integration of large-scale data centers into the electric grid, supporting workforce development at the intersection of computing and energy systems, and creating educational and research opportunities for students in emerging areas of critical national importance. This project aims to develop a unified modeling, analysis, and control framework for understanding interactions between data centers and electric power systems. The research comprises three interconnected thrusts. The first thrust develops reduced-order models of data-center power-electronic systems to characterize how internal converters, control loops, and energy-storage elements shape the load and impedance seen by the grid. The second thrust develops stochastic models of computing workloads to capture bursty, temporally correlated, and scheduler-driven compute demand. The third thrust develops system-level analysis and control methods to assess and improve grid stability under dynamic operating conditions. By linking workload dynamics, power-electronic behavior, and grid-level responses, the project seeks to determine how compute-driven demand is transformed into the effective load and impedance observed by the grid. The resulting framework will enable scalable analysis of stability margins, disturbance propagation, and control design for future power grids serving highly dynamic data center loads. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.