Archived

This content is available here for research, reference, and/or recordkeeping.

Abstract

The rapid advancement and widespread integration of artificial intelligence (AI) is driving demand for unprecedented deployment of power-intensive computational infrastructure, including multi-megawatt data centers with the potential for facilities with gigawatt-scale capacity in the near future. In this paper, load growth projections for the US are reviewed, and an example energy dispatch solution considering a mixed energy portfolio with flexible, renewable, distributed, and load-based generation is employed. The brief technology review included in the paper covers aspects of electric power, cooling, and computational infrastructures. The concept of a data center digital twin for transient load, grid interaction, and hybrid energy dispatch studies is described and proposed for implementation using a power hardware-in-the-loop test bench. For the transients associated with the typical data center loads, a generative adversarial neural (GAN) model is proposed and employed to synthetically produce GPU load profiles at the rack level, which are aggregated to emulate large-scale data center behavior.

Document Type

Conference Proceeding

Publication Date

Fall 10-2025

Notes/Citation Information

Fischer, G. M., Alden, R. E., Lewis, D. D., Patrick, A., and Ionel, D. M., "Data Center Developments for Flexible Generation Dispatch, Advanced Infrastructure, and Ultra-Fast Digital Twins," Proceedings, IEEE International Conference on Renewable Energy Research Applications (ICRERA), Vienna, Austria, doi: 10.1109/ICRERA66237.2025.11283760, 6p (Oct 2025)

Digital Object Identifier (DOI)

10.1109/ICRERA66237.2025.11283760

Download

Share

COinS