Probabilistic Risk Assessment for Integrated Energy Systems with Data Centers
Background & Objectives
Integrated energy systems (IES) couple multiple energy sources, such as nuclear, fossil fuels, and renewables, with interconnected energy networks, including electricity, natural gas, hydrogen, and heat, to improve energy efficiency and enhance supply reliability and resilience. Understanding and quantifying risks in IES is essential, as integrating multiple energy networks and technologies can introduce complex operational interdependencies spanning a broad range of time and spatial scales. Conventional risk analyses focusing on a single energy network (e.g., a power grid) do not fully capture these interdependencies.
The growing energy demand from data centers adds a new dimension to this challenge. Data centers are not only power consumers but also active participants when co-located with on-site power generation facilities, such as nuclear reactors and gas turbines. Beyond electricity, data center operations depend on multiple energy carriers, such as natural gas for backup power generation and thermal energy for cooling. Analyzing the energy supply risk of data centers, therefore, requires an IES-level perspective that explicitly accounts for interdependencies among multiple energy networks and on-site generation.
This research develops a probabilistic risk assessment (PRA) framework for multi-network IES risk assessment, with explicit treatment of data center loads and their onsite energy infrastructure, to support risk-informed planning and operation of integrated energy systems with large-scale data centers connected.
Our Approach
This research project is organized based on three main tasks:
- Conduct a systematic literature review to assess the state of the art in IES risk analysis, identify key research gaps in existing methods, and define the scope and requirements for the probabilistic IES simulation framework, particularly when applied to IES with data centers connected.
- Develop a probabilistic simulation framework for multi-network IES risk assessment. Each energy network, including electricity, natural gas, and others, is modeled as a distinct simulation module based on physical governing equations and operational constraints. Uncertainty in supply, demand, and component availability is propagated using a Monte Carlo simulation. The sampling process is structured using dynamic event tree analysis to improve the interpretability of risk outputs in relation to initiators and event sequences.
- Extend the framework to explicitly model data center loads connected to multiple energy networks, including their onsite energy infrastructure, such as backup batteries and gas turbines or nuclear reactors. This extension supports risk assessment of the compound energy supply risk faced by large-scale data centers that source power from both grid-connected sources and co-located generation facilities.
Ongoing work focuses on coupling the gas network module with an electric power network model and extending the framework to incorporate data center loads and onsite generation.
Impact and Broader Applicability
- Advance the understanding of energy supply risk arising from the connection of hyperscale data centers, providing quantitative insights into the interdependencies among multiple energy networks.
- Demonstrate the benefits of nuclear power quantitatively, both as a grid-connected baseload source and as a co-located onsite generation facility, in terms of enhancing the reliability and stability of energy networks with data centers connected. This will help provide an overarching, system-level perspective on nuclear power's role in the evolving national energy landscape.
- Identify practical risk reduction measures for data center energy infrastructure, including optimal configurations of onsite generation and storage assets.
References
E. Stouffer, A. Kalantari, and T. Sakurahara, "Probabilistic Risk Assessment for Multi-Network Integrated Energy Systems: Gap Analysis and Preliminary Results," Proceedings of the 18th Probabilistic Safety Assessment & Management Conference (PSAM18), Pittsburgh, PA, 2026.