The Illusion of Precision in Energy Forecasting

“Everyone has a plan until they get punched in the mouth.”

Mike Tyson was talking about boxing, but he could just as easily have been talking about energy infrastructure.

Numbers carry enormous persuasive power in climate and energy debates. Cost projections are cited confidently in dollars per megawatt-hour. Deployment timelines are presented with calendar-year precision. Entire policy systems are built around models that forecast which technologies will dominate decades from now.

The problem is not that modeling is flawed, but that modeling is misunderstood to be certain. The risk is not uncertainty itself, but the illusion of precision.

Energy models are powerful tools for exploring possibilities, but they are frequently treated as prescient instruments that deliver singular answers about the future. In reality, they produce best estimates that are conditioned on assumptions. They map possibility space so engineers can prepare for the punch, because in complex infrastructure, the punch always comes.

As electricity systems grow more complex and climate ambitions more urgent, misunderstanding models to be precise carries consequences.

Models Are About Ranges, Not Certainty

Electricity grids are among the most complex engineered systems in society. They must regulate supply and demand in real time while integrating physical infrastructure, market rules, fuel logistics, weather variability, and human behavior. The large infrastructure projects that comprise electricity grids – such as nuclear power plants – are themselves feats of engineering that require billions of dollars in capital and coordination across thousands of skilled workers and suppliers.

At their best, models illustrate how variables interact. Change an assumption, such as fuel prices, learning rates, or demand growth, and the outputs shift. Understanding how input uncertainty affects possible outcomes reveals which outcomes are robust, and which depend on fragile assumptions.

In simple systems, small changes in assumptions produce manageable shifts in outcomes. Modern electricity infrastructure is anything but simple. Uncertainty compounds when models scale to national or global systems, and the more variables interact, the less meaningful a single expected value becomes.

Projecting that a technology will cost a certain amount in 2040 requires assumptions about capital costs, financing conditions, policy stability, supply chains, and deployment rates over decades. Each assumption carries its own uncertainty distribution. When those distributions interact, the resulting range of outcomes can be wide, even if the central estimate appears precise.

This challenge is particularly acute for first-of-a-kind advanced nuclear reactors. Estimating the 100,000-hour failure rate of a component that has never been built requires predicting not only uncertainty, but also the uncertainty in that uncertainty.

This does not invalidate modeling, but rather clarifies that models are best interpreted as maps of probability space rather than forecasts of inevitability.

How Nuclear Engineering Designs for “the Punch”

Uncertainty is not abstract in high-risk engineering disciplines. Analyzing and quantifying it is standard practice in fields such as aviation, structural design, and nuclear safety.

In nuclear safety analysis, Monte Carlo simulations and probabilistic methods predict not only what outcome is most likely, but under what combinations of failures the system breaks. Model uncertainty is quantified, margins of error are bounded, and multiple failure modes are considered simultaneously.

Probabilistic Risk Assessment (PRA) models how nuclear reactor safety systems interact and how combinations of unlikely failures could lead to reactor core damage. Instead of producing a single deterministic answer, PRA generates probability distributions of outcomes by sampling across initiating events, component reliability, system interactions, and human error. The result is not certainty – but a structured understanding of risk.

In nuclear criticality safety, sensitivity coefficients and “similarity” analysis enable experimentalists to design better benchmark experiments by quantifying whether uncertainty affects potential experiments in the same way it would affect full-scale nuclear reactors. These methods are used to design essentially all critical experiments in the US and abroad; in fact, these methods can be used to verify that the recent Valar Atomics criticality milestone – which used fuel fabricated in 1987 for Los Alamos National Laboratory’s Compact Nuclear Power Source (CNPS) – accurately represents Valar’s proposed reactor fuel. These methods aim not just to quantify uncertainty, but to predict if uncertainty will impact different systems in the same way.

This uncertainty mindset is standard in high-risk engineering projects for a reason. Uncertainty is not a complication to hide, but a variable to explore. It would be easy if everything went according to plan, but in reality complex systems fail because of the exceptions.

The Cost of Deterministic Narratives

The recently constructed Vogtle Unit 3 and 4 AP1000 reactors illustrate the stakes of neglecting uncertainty.

Early budgeting and decision-making did not rely heavily on probabilistic modeling. However, after delays and cost overruns became evident, probabilistic methods became critical for estimating the reactor schedule completion probability and projected cost-to-completion. The second of the two AP1000’s – Unit 4 – benefited from the lessons learned and was ultimately 20-40 percent cheaper to construct than the first unit.

The lesson learned from Vogtle is not that modeling failed, but that models are most useful when they illustrate risk ranges early – before reality forces the lesson. When forecasts are treated as promises rather than scenarios, planning can crystalize around assumptions that reality later disproves. When operating complex systems under high stakes, humility is an operational requirement.

Asking Better Questions

The questions we ask energy models matter.

Instead of asking which technology will be cheapest in 2040, we might ask how sensitive that comparison is to financing conditions. Instead of asking whether a pathway will dominate, we might ask under what assumptions it succeeds and how plausible those assumptions are.

These questions do not produce neat headlines, but they do produce more resilient planning.

As electricity demand rises, driven by electrification, data centers, and climate adaptation itself, the cost of excessive confidence grows. Infrastructure investments today will shape emissions trajectories for decades. Decisions made on narrow readings of broad uncertainty can lock in fragility. Systems must function under stress; construction must proceed under stress; supply chains must be reliable under stress. Probabilistic thinking helps to anticipate these stresses.

Uncertainty as Responsibility

Accepting the limits of modeling is not an argument for paralysis. It is an argument for discipline.

Climate and energy transitions are engineering challenges as much as political ones. They require long-lived infrastructure, resilient systems, and adaptive capacity. They benefit from explicit ranges, sensitivity analyses, varied scenarios, and recognition of compounding risk.

Models are essential, but they are most powerful when treated as guides to complexity rather than guarantees of certainty. In high-risk systems, confidence should scale with comprehension rather than optimism.

As the energy transition accelerates, the most responsible viewpoint in public discourse may not be louder promises, but clearer recognition of what we know, what we do not know, and how the gap between the two shapes the path forward.

About the Author

Dr. Christopher Perfetti is an Associate Professor of Nuclear Engineering at the University of New Mexico and an expert in reactor physics and uncertainty quantification. Previously an R&D scientist at Oak Ridge National Laboratory, he developed advanced sensitivity and uncertainty tools for the SCALE code package, including CE TSUNAMI-3D. He is an award-winning leader within the American Nuclear Society and a Nuclear News 40 Under 40 honoree.

Dr. Christopher Perfetti

Associate Professor of Nuclear Engineering, University of New Mexico

Author profile LinkedIn