Cost and time overruns of the Areva EPR reactors at Olkiluoto in Finland and Flamanville in France are seldom out of the energy news. Olkiluoto began construction in 2005 with planned grid connection in 2010. The original build cost of €3billion has risen to €8.5 billion. And the grid connection has been pushed out to 2018 – 8 years late (13 years construction time) and €5.5 billion over budget.
So how long should it take to build a nuclear reactor? 5, 10 or 15 years? The answers are below the fold.
The Areva EPR being built in Flamanville France has experienced similar delays. With an original planned construction time of 54 months (5.5 years) and budget of €3.3 billion the project is now heading for 11 years construction and a price tag of €10.5 billion (all preceding stats from Wikipedia).
This post has its origins in my previous post on the Global Nuclear Power Snapshot. Respected Russian commenter Syndroma asked a question about the age of the global nuclear fleet and referred to the IAEA PRIS database and then he answered his own questions by posting a couple of charts. Extracting the data from the IAEA pdf is not entirely straight forward and Syndroma is thanked for doing this for me.
The PRIS data base contains information on 441 operational reactors including the date of build start and the date of grid connection allowing us to calculate the construction time. Sorting the data on construction time produces the distribution shown in Figure 1.
Figure 1 The distribution of nuclear power station build times from the IAEA PRIS database.
I was surprised to see that 374 out of 441 reactors were built in 10 or less than 10 years. There is a tail of 15% that have taken longer to build. The world record is 33 years for the Atucha-2 reactor in Argentina where construction began in 1981 and was grid connected in 2014. I can only assume that construction was halted for a large number of years.
At the other end of the scale, 18 reactors were completed in 3 years! 12 of those in Japan, 3 in the USA, 2 in Russia and 1 in Switzerland. These are a mixture of boiling water and pressurised water reactors. Clearly, it does not need to take forever to build new reactors given good supply chain, expertise and engineering protocols. The mean construction time of 441 reactors in use today was 7.5 years.
There is often talk in nuclear circles that over-regulation has led to increased costs and build time. If this were the case we may expect to see an increase in construction time over time. Figure 2 shows the construction time cross plotted with reactor age.
Figure 2 The trend of increasing construction time with decreasing age began 45 years ago and there is no clear evidence that reactors are taking longer to build today than they did 40 years ago.
Figure 2 displays 2 trends. The vertical trend shows no correlation between reactor age and construction time. But the diagonal trend does show a correlation between construction time and age in some instances. But this is inconclusive since those reactors that took a long time to build (lower right quadrant) are by definition young since they are only recently completed.
Regulation differs between the OECD and non-OECD countries and so I applied one final test to see if over-regulation in the OECD led to long construction times.
Figure 3 There is perhaps limited evidence that in some cases regulation in the OECD has led to longer construction times. But most of the young OECD reactors, <15 years old, have short construction times of around 5 years.
The OECD group defines both trends and there is some evidence here that in some cases younger reactors have taken longer to build. The record holders in the OECD are Watts Bar 1 in the USA at 23 years, Dungeness B2 and B1 in the UK at 20 and 18 years respectively.
Figure 4 The non-OECD countries have a group of 9 problem reactors (circled) that are listed in Figure 5. Many recent non-OECD reactors (<15 years old, lower left quadrant) have build times between 5 and 10 years, in fact longer than the OECD.
The non-OECD can be divided into two groups. One where construction time and age are not correlated and a problem group of 9 reactors that took forever to build (circled). These are listed in figure 5 and commenters are invited to elaborate on the causes behind the long delays experienced in these projects. I’d note that 6 of the 9 are the Russian VVER V model.
Figure 5 9 non-OECD reactors with long construction times. Click image for a large readable version.
So what has gone wrong at Olkiluoto and Flamanville? Nothing really apart from Areva being hopelessly optimistic in their original forecasts of build time (5 years) and costs. The average time taken to build 441 reactors operational today was 7.5 years. For Areva to believe they could build first of type Gen 3 EPR reactors in 5 years was optimistic to say the least. The time and cost over runs at Olkiluoto and Flamanville are only bad compared with the original plan but are not yet catastrophic in absolute terms. But let’s hope they get Hinkley Point C down to the 7.5 year mean.