Case Study: The mystery of the missing learning curve


Problem Statement: In 1964, the Atomic Energy Commission (AEC) which is the predecessor of the Nucelar Regulatory Commission (NRC) hypotesized that a learning curve would exist for nuclear plant construction, and that the cost would gradually reduce to less than $1000/kWe (2011 USD). Figure S1 shows the hypothesized curve, and shows that this learning curve never materialized based on the the actual costs which in some cases were up to 8 times what was predicted. 

Figure S1: Shows hypothesized learning curve and actual costs, along with major incidents that occurred in the nuclear industry
Major Finding: Diseconomies of scale exist in nuclear power plant construction


To determine the reasons for the absence of learning, we performed a historical analysis of all nuclear power plants constructed in the US. We found that a major cause of cost overruns is estimation errors in the initial cost and schedule planning process. One systemic issue is the lack of recognition of the diseconomies of scale that exist in nuclear power construction. Large plants are more complex and hence cost more on a cost per kWe basis. This is counterintuitive, and prior and current estimation methods assume economies of scale which leads to errors in estimates.

Figure S2: Shows relationship between cost overruns and plant size
Other Findings....

We found several systemic issues with the processes for cost and schedule estimation on large nuclear projects. These issues include a lack of correlation between cost and schedule estimates, and faulty assumptions on the effects of experience. These findings and others are documented in peer refereed publications that we authored.