Evaluating Grid Resilience in the Era of Ever-Increasing Data Centers

arXiv:2607.06958 · eess.SY, cs.SY · Submitted 2026-07-08 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: We're looking at the paper: "Evaluating Grid Resilience in the Era of Ever-Increasing Data Centers".

Dev: Here's the gist of the paper. The gist: Data center capacity growth substantially increases both system-level and data-center-bus unserved energy under transmission-constrained contingencies, and temporally concentrated data center demand can amplify resilience impacts even without increasing total energy consumption.

Dev: The framework extends a validated multi-time-step DCOPF resilience model developed in

4: (Page 2).

Rosa: At each time step t, the model determines available generation, battery charging and discharging, power flows, and emergency unserved demand while considering projected load and disruption conditions over a finite decision horizon (Page 4).

Title and authors: Rosa: So, we're looking at this paper called "Evaluating Grid Resilience in the Era of Ever-Increasing Data Centers." It seems like they’re taking a system that models power flow and adding in these massive data centers to see how it holds up when things go wrong.

Dev: Yeah, the title itself sets the stage—it's all about how increasing data center demand messes with grid resilience when we have disruptions. The authors are Du, Ardiles-Cruze, and Mohammadi from the University of Texas at Austin and Air Force Research Labs.

Taro: I’m curious if this study is just looking at big numbers on paper or if they actually tested how these data center loads behave when the world misbehaves.

Rosa: Well, based on what I see here, they are using a specific setup—an IEEE thirty-bus system with flexible resources—to replace a regular load at a vulnerable bus with this new data center load to test the impact of capacity growth under different types of transmission stress <ref:2607.06958#pg1,an IEEE 30-bus system with flexible resources>.

Dev: They are specifically interested in how data centers, especially those driven by AI workloads, might introduce new kinds of temporal demand characteristics that we haven't seen before in standard dispatch models.

Taro: So they’re not just looking at the total size of the data center load, but how it shifts over time when a disruption hits.

The paper's summary: Rosa: The summary says they took their existing multi-time-step DCOPF framework and used it to see how aggregated data center demand impacts unserved energy under generator and transmission derating, coordinating flexible resources like batteries.

Dev: That’s the core idea—they built on a model that already knows how power flows and resources react, but now they’re injecting this aggregated data center load as a growing reliability concern because its size and operational characteristics can affect transmission planning and real-time system operation.

Taro: They are using an energy-matched comparison to separate the effects of the data center's load profile from just how much its total capacity is growing, which is important because those two things can interact in complex ways.

Rosa: And they introduce this dispatch-scale coincident-demand case to specifically test what happens when that high growth demand overlaps with a grid disruption, which addresses those new temporal demand characteristics we talked about earlier.

Dev: Essentially, the paper is trying to figure out if having these data centers ramp up their power consumption right when the grid is already stressed makes things significantly worse than just having a larger fixed load.

The paper's improvements: Taro: I noticed they specifically mention extending a validated multi-time-step DCOPF framework with an aggregated data center load representation at a contingency-exposed node, which is the first major technical improvement mentioned.

Rosa: Exactly, and then they develop this energy-matched comparison to evaluate the effects of data center load profiles and capacity growth under different derating scenarios, like generator derating or line derating.

Dev: And on top of that, they introduce a dispatch-scale coincident-demand case to specifically test what happens when that high growth demand overlaps with a grid disruption, which is how they address the temporal concentration issue we discussed.

Taro: That means they’re not just looking at static loads; they’re testing dynamic stress by comparing the constant high-growth profile against a coincident-demand case during a disruption, which helps quantify the amplification of unserved energy.

Rosa: So it suggests that for resilience planning, you need to look at both how much capacity is growing and when that growth happens relative to the system stress.

Dev: That's the practical implication for us—it’s not just about installing more power; it’s about understanding the timing of that power draw during a crisis.

Conclusion: Rosa: To wrap up, this paper shows that data center capacity growth substantially increases both system-level and data-center-bus unserved energy under transmission-constrained contingencies.

Dev: And they also found that an energy-matched coincident-demand case increases total unserved energy by thirty-four point four percent under coupled derating scenarios, which really highlights the importance of temporal demand concentration in resilience assessment.

Taro: It seems like the paper concludes that an energy-matched comparison shows that the constant data center profile produces lower unserved energy than the original peak-period conventional profile in this contingency-exposed setting.

Rosa: So while capacity growth is a big problem when transmission is constrained, having a constant load might be better than fluctuating demand during those specific events.

Dev: It’s a useful tool for operators to understand that temporal concentration can materially amplify the impact of a disruption even when the total energy demand isn't changing.

Taro: Anyway, this paper, "Evaluating Grid Resilience in the Era of Ever-Increasing Data Centers," gives us a clearer picture of how these huge new loads interact with system failures.

Yuhan Du, Erika Ardiles-Cruze, Javad Mohammadi

University of Texas at Austin · Air Force Research Labs

eess.SY, cs.SY

Submitted: 2026-07-08

Updated: 2026-10-20

Comments: 8 pages, 2 figures, accepted for Dynamic Data Driven Applications Systems (DDDAS) 6th International Conference

License: http://creativecommons.org/licenses/by-sa/4.0/

Importance score: 72/100

The gist: The gist: Data center capacity growth substantially increases both system-level and data-center-bus unserved energy under transmission-constrained contingencies, and temporally concentrated data

Key concepts

System-level Unserved Energy (UEsys)
This metric measures the total amount of electricity that cannot be supplied to the entire power grid during a disruption. It shows how much the overall system fails when stressed, including impacts on generation and transmission lines.
Data Center-Bus Unserved Energy (UEDC)
This focuses specifically on energy shortages at a single bus where a data center is located. It quantifies the local impact of grid failures directly affecting that critical facility, helping operators prioritize resilience efforts at key sites.
Capacity Growth Cases
These scenarios simulate future data center expansion by increasing the baseline energy demand by 1.5x or 2x. This tests how infrastructure performs when facing larger, more demanding data centers than currently planned.
Temporal Demand Concentration
This concept examines how concentrated, high-growth data center demand behaves during a disruption. It shows that if a large spike in demand occurs simultaneously with a grid failure, the resulting energy shortage can be much worse than if the total load were spread out.

Terminology

Summary

The gist: Data center capacity growth substantially increases both system-level and data-center-bus unserved energy under transmission-constrained contingencies, and temporally concentrated data center demand can amplify resilience impacts even without increasing total energy consumption.

Methodology

  1. The framework extends a validated multi-time-step DCOPF resilience model developed in [4] (Page 2). At each time step t, the model determines available generation, battery charging and discharging, power flows, and emergency unserved demand while considering projected load and disruption conditions over a finite decision horizon (Page 4). The optimization objective is represented as minX t C G t + C B t + c S X i Si,t! (Page 4).

  2. The data center is represented as an aggregated grid-facing demand at a selected bus d (Page 4). To separate temporal load-profile effects from increased demand, the energy-matched constant data center demand (DC) is defined as L DC d = 1/NT X NT t=1 L O d,t (Page 4). Capacity-growth cases are then represented by L DC,β d,t = βL DC d, β ∈ [1.5, 2.0] (Page 4).

  3. A dispatch-scale coincident-demand sensitivity case for the high-growth data center profile is introduced where the swing proxy is defined as L SW d,t = (1 + α)L HG d, t ∈ TD, 1 − α TD NT − TD L HG d /∈ TD (Page 4).

Resilience Metrics

The paper evaluates resilience using two primary metrics: total unserved energy (UEsys) and data-center-bus unserved energy (UEDC) (Page 5). For the data center bus, the served-energy ratio is also reported as R DC = 1 − UEDC / ∆tP t Ld,t (Page 5).

Case Study Setup

The evaluation uses the IEEE 30-bus system with flexible resources from the reproduced baseline model [4] (Page 5). The simulation employs a 15-minute resolution and a 36-step decision horizon, corresponding to 9 hours, which was selected because no further reduction in unserved demand was observed beyond this for the coupled-derating case (Page 5). Bus 7 is selected as the data center location due to its exposure to contingencies (Page 5). The energy-matched constant data center demand is set to 26.283 MW, while capacity-growth cases are set to 39.424 MW and 52.565 MW, corresponding to 1.5× and 2× the energy-matched demand, respectively (Page 5). Three contingency scenarios are included: generator derating (generation capacities at buses 1 and 2 reduced by 90%), line derating (lines reduced to 5% of their nominal flow limits), and coupled derating (both deratings applied simultaneously) (Page 5).

Results and Discussion

The results show that data center capacity growth substantially increases unserved energy, particularly when transmission delivery is constrained. Under line derating, total unserved energy increases from 3.079 MWh in the DC-matched case to 22.766 MWh in the high-growth case (Page 6). Under coupled derating, total unserved energy similarly increases from 3.203 to 22.891 MWh, with data center bus unserved energy increasing from 0 to 19.665 MWh (Page 6).

To examine temporal demand concentration, the constant high-growth profile is compared with an energy-matched coincident-demand case under coupled derating (Page 7). As shown in Fig. 1, disruption-coincident high demand increases total unserved energy from 22.891 to 30.777 MWh, corresponding to a 34.4% increase (Page 7). Data center bus unserved energy increases from 19.803 to 27.688 MWh, or a 39.8% increase, while the data center served-energy ratio decreases from 95.81% to 94.15% (Page 7). This sensitivity case indicates that temporally concentrated data center demand can materially amplify contingency impacts even when total energy demand is unchanged (Page 7).

Conclusion

The paper concludes that an energy-matched comparison shows that, in the studied contingency-exposed setting, the constant data center profile produces lower unserved energy than the original peak-period conventional profile (Page 8). In contrast, data center capacity growth substantially increases system-level and data-center-bus unserved energy under transmission-constrained disruptions (Page 8). Moreover, an energy-matched coincident-demand case increases total unserved energy by 34.4% under coupled derating, highlighting the importance of temporal demand concentration in resilience assessment (Page 8). Future work will incorporate higher-fidelity data center operating models, alternative siting conditions, and mitigation strategies such as grid-interactive storage and controllable computing demand (Page 8).

References

  1. Chen, M., Gao, C., Song, M., Chen, S., Li, D.: Internet data centers participating in demand response: A comprehensive review. Renewable and Sustainable Energy Reviews 117, 109466 (2020). https://doi.org/10.1016/j.rser.2019.109466

  2. Choukse, E., Warrier, B., Heath, S., Belmont, L., Zhao, A., Khan, H.A., Harry, B., Kappel, M., Hewett, R.J., Datta, K., et al.: Power stabilization for ai training datacenters. arXiv preprint arXiv:2508.14318 (2025).

  3. Crozier, C., Liska, M.: The potential of data center energy demand to provide grid flexibility. Current Sustainable/Renewable Energy Reports 12, 12 (2025). https://doi.org/10.1007/s40518-025-00258-9

  4. Du, Y., Golan, M., Mohammadi, J.: Measuring power grid resilience in the era of flexible resources using multi-time-step dispatch models. In: 2025 57th North American Power Symposium (NAPS). IEEE (2025). https://doi.org/10.1109/NAPS66256.2025.11272383

  5. Golan, M.S., Mohammadi, J., Linkov, I.: Dimensions of resilience in power grids: Strategies for quantifying resilience. Electric Power Systems Research 255, 112736 (2026). https://doi.org/10.1016/j.epsr.2026.112736

  6. Guo, C., Luo, F., Cai, Z., Dong, Z.Y.: Integrated energy systems of data centers and smart grids: State-of-the-art and future opportunities. Applied Energy 301, 117474 (2021). https://doi.org/10.1016/j.apenergy.2021.117474

  7. Liu, L., Shen, X., Chen, Z., Sun, Q., Wennersten, R.: Optimal energy management of data center micro-grid considering computing workloads shift. IEEE Access 12, 102061–102075 (2024). https://doi.org/10.1109/ACCESS.2024.3432120

  8. Liu, Z., Wierman, A., Chen, Y., Razon, B., Chen, N.: Data center demand response: Avoiding the coincident peak via workload shifting and local generation. Performance Evaluation 70(10), 770–791 (2013). https://doi.org/10.1016/j.peva.2013.08.014

  9. North American Electric Reliability Corporation: Risk mitigation for emerging large loads: Large loads working group reliability guideline. Reliability guideline, North American Electric Reliability Corporation (May 2026), https://www.nerc.com/globalassets/our-work/guidelines/reliability/RG Risk-Mitigation-For-Emerging-Large-Loads.pdf

  10. Panteli, M., Mancarella, P.: The grid: Stronger, bigger, smarter? presenting a conceptual framework of power system resilience. IEEE Power & Energy Magazine 13(3), 58–66 (2015). https://doi.org/10.1109/MPE.2015.2397334

  11. Panteli, M., Mancarella, P.: Modeling and evaluating the resilience of critical electrical power infrastructure to extreme weather events.

Improvements for AI systems

  1. Enhanced Contingency Resilience Planning: The improved system can accurately evaluate how data center capacity growth substantially increases both system-level and data-center-bus unserved energy under transmission-constrained contingencies by incorporating dynamic load profiles rather than just static averages.

  2. Temporal Concentration Risk Mitigation: The AI system can implement a dispatch-scale coincident-demand sensitivity case to quantify how temporally concentrated data center demand can affect system-level and data-center-bus unserved energy, allowing operators to predict and mitigate amplified impacts when demand spikes during disruptions.

  3. Load Profile Optimization: Using the energy-matched comparison, the improved system can determine if a constant data center profile produces lower unserved energy than the original peak-period conventional profile at a contingency-exposed node, enabling more resilient load scheduling decisions.

Abstract

The rapid growth of artificial intelligence workloads is increasing the scale and concentration of data center demand, creating new concerns for power system resilience under disruptive events. This paper extends a validated multi-time-step DC optimal power flow framework to evaluate the impact of aggregated data center demand on contingency-induced unserved energy. Using an IEEE 30-bus system with flexible resources, we replace a conventional load at a contingency-exposed bus with an energy-matched constant data center load and examine two capacity-growth levels under generator derating, transmission line derating, and coupled derating. The results show that data center capacity growth substantially increases both system-level and data-center-bus unserved energy under transmission-constrained contingencies. Under coupled derating, the high-growth case increases total unserved energy from 3.203 MWh in the energy-matched case to 22.891 MWh. A supplementary energy-matched coincident-demand case further increases total unserved energy by 34.4%, indicating that temporally concentrated data center demand can amplify resilience impacts even without increasing total energy consumption.

Sources

Related papers