2.4 SMR Cost Estimating Challenges
2.4.1 Challenge 1: Applicability/ Availability of Existing Data
The estimation of costs is driven by both predicted and actual experience of the performance of similar technology (Woodward, 1997). One of the main challenges associated with estimating the cost of SMRs is the application of reliable data to represent future costs. Actual SMR cost data is not available, nor are detailed cost estimates for the designs under development (Shropshire, 2011). The literature review has identified a variety of sources and methods used to estimate SMR costs ranging from large NPP data, expert elicitation, to simulation and modelling.
2.4.1.1 Existing NPP Data
The large nuclear reactor fleet in the US provide the richest data source for generation costs, with the most comprehensive analysis produced by Koomey and Hultman (2007). In their analysis, Koomey & Hultman obtained capital cost, construction duration and capacity information from the database produced by Komanoff (1981) and extrapolated to 2006 rates using operations costs, including fuel costs, together with thermal efficiencies derived from The Nuclear Energy Institute, US Energy Information Administration, and IAEA data sources. Komanoff’s database of capital cost data was based on utility records produced by the Energy Information Administration, converted to 1979 dollars.
36 Both the Grubler (2010) and Boccard (2014) studies relied upon data released by the French government showing actual data, via the Court of Audit. The OCC for nuclear reactors built in the UK, France and US have been analysed to try to identify the causes of cost escalation when comparing NPPs produced in the first generation (construction start 1966-67) versus later generations (construction start 1974-77) (MacKerron, 1992). Moreira et al (2013) use the construction time to represent cost escalation of PWRs built in the 30 years before 2012.
Actual data regarding construction and operational costs are difficult to obtain, and the methods used to record actuals in different countries and different times is dependent on the accounting procedure employed. The data is subject to different methods of accounting, use of financial and inflation cost indices, and other adjustments. Analysing and comparing performance using the different sources of data available will be subject to error due to the number of assumptions needed to normalise the data.
One method for estimating novel technologies is to verify the assumptions in new estimates with historical data of costs for a similar technology (Roy, 2003). Carelli (2010) estimated the LCOE for an International Atomic Energy Agency reference design SMR using an analogy type estimate from large reactor data. Using bottom-up estimation with parametric cost estimating relationships to scale down a large NPP, Paparusso (2012) showed that SMR costs are less subject to capital costs. For an SMR, the comparative parameters usually include economies of learning, co-siting, and degree of modularity of the design relative to a large reactor (Sultan & Kattab, 1995).
2.4.1.2 Simulation and Modelling
Models can be used to understand the future cost of energy generation technologies using experience curves based on historical data (Neij, 2008). Historical experience curves are reliant on a standardised product to analyse the associated cost reductions. Scenario analysis is often used to understand the comparative energy costs with varying costs of wholesale electricity or other, technology specific costs (Kennedy, 2007). For future wholesale electricity prices, the University of Chicago (2004) study considered different future electricity demand scenarios, all based on the predictions of electricity market models.
37 There is great uncertainty as to the level of standardisation achieved historically by the nuclear industry, with potentially only France able to claim a level of learning from doing (Boccard, 2014). Different NPPs technologies, specific site related conditions, production methods, construction companies, operating strategies, operating environment and financing all contribute to the reduction in NPP product standardisation.
2.4.1.3 Expert Elicitation
Where historical data is not considered representative of future designs, or in situations where there is a lack of available historic data, expert elicitation is identified as a reasonable method of understanding future costs (Levi & Pollitt, 2015). For example, Anadón et al (2013) used expert elicitation to obtain values for the overnight construction cost for a SMR, using this to form an input into the LCOE calculation. Database cost estimates are often supplemented with expert judgement to make the data fit the new scenario (Roy, 2003).
2.4.1.4 Problems with Data Sources
There is a lack of consistent treatment of cost data and financial reporting structure. Actual data regarding construction and operational costs are also difficult to obtain, making the consistent recording, analysing and comparing of cost performance a challenge. The variability is in part due to a lack of available data or granularity of data. It is also the result of the long timescales and lack of experience in construction of new NPPs, leading to great uncertainty in the estimates produced. Reviews of historical experience, therefore, tend to rely on non-financial data, such as construction time, to interpret the likely cost of capital. The extent to which causality can be established between construction time and construction schedule is questionable. Different studies use different values for the construction time, based on the historical trend for construction around the world, reactor vendor marketing information, or based on a range of possible scenarios. Other parameters should be considered such as the reactor type, power output, and average availability over time.
Even when the LCOE estimate is based on statistically valid historical data from a nuclear build programme, there is still a high level of uncertainty with respect to the future cost of construction, operation and decommissioning. Selecting the most
38 relevant data for estimating the cost of SMRs is dependent on many assumptions. Determining which assumptions are representative of the SMR relies on data that may not be statistically valid and may therefore be highly uncertain. An understanding of the sources of data and assumptions used to calculate the estimate is required to assess the reliability of results. Due to the large number of direct and external influencing variables, historical experience may not be a valid benchmark to identify the future LCOE of the SMR.