Data-driven Assessment of Magnetic Resonance Imaging (MRI) Utilization and Aggregate Electrical Demand in U.S....

Magnetic resonance imaging (MRI) systems are among the most energy-intensive assets in healthcare facilities and are typically designed using conservative electrical criteria that may not reflect actual operational behavior. This study develops a data-driven framework to characterize MRI energy use, operational patterns, and aggregate electrical loading, with the objective of identifying opportunities to safely reduce electrical infrastructure oversizing.

One-minute-interval power data were collected from 13 U.S. hospitals, yielding 965 detected MRI events. Events were clustered into six groups using k-means clustering. A first-order Markov chain model was employed to evaluate steady-state transitions among clusters. Aggregate loading conditions were examined through Monte Carlo simulations by superposing two to ten independent MRI events and computing probabilities of exceedance (PoE).

Clustering results showed that MRI operations are dominated by short- to mid-duration scans. Markov analysis revealed high self-transition probabilities, suggesting temporal stability in event durations. Although short-duration current spikes produced nonzero PoEs, no system failures were observed. Under a 1% PoE criterion, electrical infrastructure ratings could be reduced by up to 45% for ten units operating simultaneously.

Predicting event duration transitions supports the development of more effective scheduling models. Probabilistic analysis of superposed MRI events suggested potential reductions in electrical capacity requirements under modeled exceedance criteria, supporting more energy-efficient and cost-effective healthcare facility design.