Frequency-Dependent Dynamic Analysis of a Prismatic HTGR Power System for Data Center Applications
Year: 2026
Type: preprint
Abstract: The rapid growth of hyperscale data centers (DCs) has increased interest in nuclear energy as a reliable low-carbon power source. However, DC demand can exhibit multi-timescale variations, and their impact on nuclear power systems remains largely unexplored. This study investigates the frequency-dependent dynamic characteristics of a prismatic high-temperature gas-cooled reactor coupled to a steam Rankine cycle using an OpenModelica dynamic model. Systematic sinusoidal load perturbations were first applied over a range of amplitudes and periods to identify dynamically sensitive demand characteristics. The resulting characteristics were then examined against a year-long facility-aggregated DC demand profile and a reconstructed workload-resolved profile based on measured AI-workload GPU power traces. The sinusoidal analysis identified an intermediate time-scale region that produced the largest plant deviations, whereas faster fluctuations were attenuated and slower variations were effectively regulated. Consistent with this behavior, the facility-aggregated profile was successfully accommodated despite repeated demand variations between approximately 15% and 100% of nominal power because its dominant variations occur over long time scales. In contrast, retaining faster workload-driven fluctuations in the workload-resolved profile produced substantially larger plant responses. Frequency-band analysis further confirmed that intermediate-frequency components contributed more strongly to these responses than higher-frequency components. For example, under PI control in the energy-matched comparison, the 95th-percentile steam-temperature deviation decreased from 7.2C to 5.5C after high-frequency attenuation and to 3.6C after intermediate-frequency attenuation. These results indicate that the spectral distribution of DC demand is an important consideration when coordinating nuclear generation with flexible resources.
Source: Engineering Archive (engrXiv)
DOI: 10.31224/8817
Authors: Wooseong Park(Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft), W. Reed Kendrick(Methodology, Writing – review & editing), Emilio Baglietto(Funding acquisition, Project administration, Writing – review & editing), Dean Price(Conceptualization, Supervision, Writing – review & editing)
Institution: Massachusetts Institute of Technology
Language: English
Subfield: Nuclear Energy and Engineering
Field: Energy
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7233760143
