Energy systems are incredibly complex, incorporating a dizzying array of power generators, distribution technologies and end-users; analyzing how all of those variables will change in the future poses challenges for long-term planning. A new method improves the computational modeling of these systems, giving policymakers new insights into which variables have the biggest impact and merit extra attention.
To learn more about this advance and why it matters, we talked to Anderson de Queiroz, co-author of a paper on the discovery and an associate professor of civil, construction and environmental engineering at NC State. The paper, "A framework for global sensitivity analysis in long-term energy systems planning using optimal transport," is published open access in the journal Energy.
The Abstract: Threshold question - when we talk about energy systems and energy systems planning, what are we talking about? What is an energy system?
Anderson de Queiroz: You can think of an energy system as the full supply chain that delivers energy to people, buildings, transportation and industry. It includes resources (such as wind, solar, coal or natural gas), conversion technologies (like turbines, generators and photovoltaic panels), transmission networks, storage and end-use demands. And all of these things are connected through technical, economic and policy rules. Planning an energy system means deciding what to build, when, and where, and how to operate it over time so that we can meet society's energy needs reliably, affordably, and sustainably.
TA: You and your co-authors on this paper are experts in modeling energy systems. What are these models, and how are they used in the context of energy systems?
de Queiroz: They're optimization models that can be used to search for the least expensive ways to build, maintain and operate energy systems in order to meet energy demand while complying with existing laws and regulations. In technical terms, we would say the models identify "least-cost investment plan/operation pathways subject to engineering and policy constraints."
Analysts use these models to run "what-if" scenarios that account for variables such as fuel prices, technology costs and carbon policies. We can test how the system would build and operate under each possible scenario. The TEMOA model (Tools for Energy Model Optimization and Analysis) is a well-known open-source example, developed here at NC State, and used at regional, national, and multi-regional scales by researchers and analysts across the globe.
TA: What model were you working with for this project, and what was the challenge or problem you were setting out to address?
de Queiroz: This paper applies a sensitivity-analysis framework to energy systems optimization models. In terms of the challenge, long-term planning models have many uncertain inputs - for example, it's impossible to know what technology costs, resource quality, demand growth and policy will be in the future. We developed a framework that can be used to understand which of these inputs actually drive the largest impacts on outputs such as energy costs, capacity build-out (such as building new power plants, transmission lines, etc.) and energy mix (how much power a system derives from different sources). Decision-makers can use this to identify which areas of uncertainty matter most.
TA: How can these findings make energy system optimization models more useful?
de Queiroz: In general terms I can think of three ways:
1) We can design more useful scenarios. If we know which variables are most important, we can focus on the inputs that dominate outcomes.
2) Targeted data collection. If electricity demand, or natural gas CAPEX or commodity prices dominate risk, we can focus on collecting better data or more detailed industry forecasts there.
3) More transparent decisions. You can better explain to non-experts why you have confidence in a model's findings about a given plan, which builds trust in both the model and the plan itself.
TA: This paper looks at energy systems on a global scale, but it also looks at Italy specifically. Why is that?
de Queiroz: You need a concrete system to demonstrate the method. Back in 2023, we started a collaboration with the Polytechnic University of Turin in the area of long-term energy systems planning. Matteo Nicoli, a Ph.D. student at the time, came to NC State to work with me in this field, and together we developed an analysis of the Italian system using TEMOA. Matteo has since completed his Ph.D. - developing methodologies for long-term energy system planning, with a case study on Italy's energy system. I had the pleasure of serving as his co-advisor with Laura Savoldi.
Italy offers diverse resources, ambitious policy targets and non-trivial technology trade-offs - making it an excellent case study to show how the modeling approach outlined in our paper can be scaled to a full national plan using TEMOA. When we did the work, there were significant concerns around Italy's supply of natural gas from Russia, creating deep uncertainty about energy costs and the country's long-term strategy.
Essentially, Italy serves as an excellent case study for how we can improve energy systems optimization models to garner key insights into systems at the national scale.
TA: Can this approach also be used by other countries? What about smaller entities, such as states?
de Queiroz: Yes, absolutely. The framework is model-agnostic enough to port to other TEMOA datasets (countries, states or regions) and, conceptually, to other long-term energy systems planning models. The outputs people care about (cost, capacity mix, emissions) are the same; you just plug in the relevant geography and assumptions.
TA: What's next? How does this work move us forward?
de Queiroz: There are several directions going forward, but I'll highlight two that I believe are particularly important today.
First, we can use what we've learned about which inputs matter most to design robust strategies for improving energy systems. Instead of just knowing that certain parameters drive costs or emissions, we can use those insights to better test how different policies or investment pathways perform across different scenarios. This will allow planners and analysts to develop adaptive and resilient portfolios, essentially identifying decisions that still work well even when the future doesn't unfold exactly as we expect.
Second, we can integrate this new approach with machine learning surrogate models. Global sensitivity coupled with energy systems planning analysis can be computationally demanding for large, detailed systems. Coupling TEMOA with machine-learning-based surrogate models (e.g., neural networks or support vector machines) could accelerate exploration across thousands of uncertain scenarios. These surrogates would approximate the optimization model's responses, allowing analysts to quickly map sensitivities and robustness across a much broader uncertainty space.
