Uncertainties in estimating production costs of future nuclear technologies: A model-based analysis of small modular reactors

Abstract

Predicting future costs of technologies not yet developed is a complex exercise that includes many uncertain parameters and functional forms. In that context, small modular reactor (SMR) concepts that are in a rather early development stage claim to have cost advantages through learning effects, standardized design, modularization, co-siting economies, and other factors, such as better time-to-market even though they exhibit negative economies of scale in their construction costs due to their lower power output compared to conventional nuclear reactors. In this paper, we compare two different approaches from production theory and show that they have a theoretically equal structure. In the second step, we apply these approaches to estimate a range of potential construction costs for 15 SMR projects for which sufficient data is available. These include water cooled, high temperature, and fast neutron spectrum reactors. We then apply the Monte Carlo method to benchmark the cost projections assumed by the manufacturers by varying the investment costs, the weighted average cost of capital, the capacity factor, and the wholesale electricity price in simulations of the net present value (NPV) and the levelized cost of electricity (LCOE). We also test whether the differences between the manufacturer estimates and ours differ between technology families of SMR concepts and apply a sensitivity analysis. Here we contribute to an intensifying debate in the literature on the economics and finance of SMR concepts. The Monte Carlo analysis suggests a broad range of NPVs and LCOEs: Surprisingly, the lowest LCOE is calculated for a helium-cooled high-temperature reactor, whereas all of the light water reactors feature higher LCOEs. None of the tested concepts is able to compete economically with existing renewable technologies, not even when taking their variability and necessary system integration costs into account. The numerical results also confirm the importance of the choice of production theory and parameters. We conclude that any technology foresight has to take as much of the case specifics into account, including technological and institutional specifics; this also holds for SMR concepts.