Explaining large observed variation in construction cost of nuclear power plants through correlated random variables

Abstract

The high overnight capital cost (OCC), as well as the large delays and cost escalation during construction, make nuclear reactors unattractive for investors. The history of nuclear power plants construction in the US shows the necessity to properly estimate construction costs and cost uncertainty and contingency. Uncertainties during construction can be classified as “known unknowns” and “unknown unknowns”. Project managers use tools to describe “known unknowns”, while “unknown unknowns”, due to their inherent nature, are not knowable. In this work, the cost trends in the various countries with nuclear power plants are analyzed. For the US data, the costs and the construction schedule of a four-loop PWR (PWR12) are also described in detail. A deterministic methodology, called EVAL, was developed to describe the construction of a nuclear power and then applied to the Westinghouse SMR. EVAL is based on a methodological approach that can evaluate construction cost for an entire Nuclear Power Plant (NPP). EVAL was applied to assess and compare different construction strategies for the Westinghouse Small Modular Reactor (WEC-SMR) nuclear island and was used to demonstrate and quantify the benefits of modularization. For this NPP design, modularization allows a 42% decrease in TCIC as compared to standard construction techniques (stick-built construction). EVAL was used to evaluate the effect of several decision variables on TCIC through sensitivity analyses. Specifically, the effect on construction costs of the discount rate, the size of the on-site assembly area, the use of different welding technologies, and testing was evaluated for the nuclear island of the WEC-SMR. A methodology to perform stochastic analyses through Iman-Conover method to account for correlation between costs and activities is also presented. A probabilistic assessment is then performed for the construction of a fully-modularized SMR and a stick-built PWR12. The results show an improved prediction capability of TCIC uncertainty as correlations between variables are taken into account. The inputs of the model are then modified to be consistent with the cost history in the US for PWR12. The trend in the US before 1979 is used to adjust the model inputs to describe a stable nuclear era. The trend after 1979 is used to quantify, a posteriori, the impact of “unknown unknowns”, representing regulatory changes during construction, resulting in cost and cost uncertainty increase. With the inputs derived from the pre-1979 data, the TCIC mean value for the PWR12-BE is $2.5 B, with a contingency of $995.5 M, which corresponds to 39.8% of the TCIC mean. Similar results were obtained for the SMR, where cost contingency is 42.0% of the TCIC expected value. Regarding the project duration, the SMR relative standard deviation is 9.5%, 10% lower than that of the PWR12-BE. If the unknown unknowns are taken into account, the PWR12-BE cost contingency is 128% of the TCIC mean derived for the pre-1979 case. For the SMR, the cost contingency relative to the TCIC mean is 53.3%, higher than that of the PWR12-BE. However, the construction time relative standard deviation for the SMR is 11.5%, about half than that of the PWR12-BE (23.2%). The analysis shows that the adoption of modular construction does not decrease OCC uncertainty with respect to stick-built construction, while it has a positive impact on the construction time uncertainty.