One of the topics in my recent discussion with Bent Flyvbjerg (Linkedin, Twitter), probably the leading global academic in megaproject failures and successes, someone who has consulted to over 100 projects costing more than a billion USD, was the optimization of modularity in scaling. The discussion point was about small modular nuclear reactors vs wind and solar energy.
The graphic below is from Flyvbjerg’s upcoming book, co-authored with Dan Gardner, How Big Things Get Done: The Surprising Factors That Determine the Fate of Every Project, from Home Renovations to Space Exploration and Everything In Between (strongly recommended for energy developers, institutional investors and policy makers). It’s assembled from the 16,000 projects Flyvbjerg and his team have gathered into their dataset of megaprojects. The black vertical line is the dividing line between projects that typically are delivered on time and budget vs those that typically aren’t, with the extremes showing which projects are best (solar and wind generation construction) and worst (all things nuclear and Olympics).
This is, of course, the problem small modular reactors (SMRS) are attempting to address: nuclear projects go over budget and schedule, especially in recent decades in the west.
As I said to Flyvbjerg, there are conditions under which GW-scale nuclear generation programs can achieve reasonable results, but they are difficult to achieve. Those conditions are that a major jurisdiction has to commit to nuclear generation as a national strategy, typically aligned with nuclear weapons. The country has to commit to dozens of reactors. They have to build them out in 30 years or less. They have to use the identical technology in each reactor with no bespoke engineering or innovations.
The combination of circumstances existed briefly in the US, France, and South Korea before withering into dust. They avoid the fat-tailed risks associated with nuclear generation ‘innovation’, avoid bespoke engineering, allow easy budgetary management from slush funds and the like, overriding of local concerns and regulatory variances from the national level, sharing of learnings from site to site, and creation of the very skilled, certified and security-cleared master builders and construction teams who know how to build the reactors quickly. It is possible to build nuclear reasonably quickly and at a reasonable cost, it’s just extraordinarily unlikely. Flyvbjerg’s nuclear project data, which includes US, French, and South Korean deployments, shows that it is still much harder to get good results than just building wind and solar.
The industry tried standardization during the ‘nuclear renaissance of the 2000s’, with the AP1000 line. It was supposed to do most of what small modular reactors are supposed to do. It was supposed to be standardized, have manufacturable components that could be shipped to sites and assembled, and be passively safe. Examples of projects with the AP1000 include the Summers and Vogtle plants, both of which went vastly over budget and schedule, one of which was killed outright and the other of which is limping along to potential grid-connection in the next year or two. The failure of the AP1000 was one of the contributing factors that led to the bankruptcy of Toshiba Westinghouse and its subsequent purchase by Brookfield for decommissioning revenue.
At the time it wasn’t clear to the many, many energy analysts who didn’t understand modularity, parallelization, manufacturability, and global supply chains that wind and solar prices would plunge. And it wasn’t clear to more people that they wouldn’t impact grid reliability. Well, the data on that is in. Wind and solar have plummeted in costs and grids with higher penetration of renewables are actually more reliable than coal, gas or nuclear heavy grids using industry standard metrics for outages per customer per year. Germany and Denmark, for example, see half or more of their annual electrical generation from renewables and have outages averaging around 13 minutes per customer per year, compared to over an hour for neighboring, nuclear-dominant France, two hours for the US and Canada, and four hours for coal-heavy Poland.
That’s not because renewables are magic, by the way, it’s just that forward thinking grid strategists and managers tend to favor renewables and also build reliable grids and markets that make them work well. Oh, and Germany’s wholesale electricity rates are among the lowest in Europe, so don’t think that this costs a lot of money.
Small modular reactors are the nuclear industry’s next big hope, because no government can gain a mandate to build dozens of GW-scale reactors any more. They keep looking at examples like Hinkley, Flamanville, and other EPR sites, the close to a trillion USD impacts to Japan’s economy due to Fukushima, and the mounting price tags and durations of nuclear decommissioning, and then look at wind and solar’s proven reliability and low prices and having trouble, even in the most nuclear-committed countries, gaining sufficient political support for what works. And so the occasional reactor that gets green lit is inevitably a failed and troubled project, resulting in very high wholesale cost electricity.
The promise of SMRs is that they will be standardized, smaller reactors that can be manufactured in central locations, shipped to sites, and assembled on site. No bespoke engineering. This has some merit, but as noted, it was mostly the promise of the AP1000 as well. The only actually different idea is that they’ll be much smaller per reactor than the AP1000, from 50-300 MW instead of 1000 MW.
I’m on record with my opinion on SMRs, and my assessment has gained surprising global attention. The original article in CleanTechnica has been updated in Illuminem, published in a (minor, low-impact factor) peer-reviewed journal, included in a German clean technology engineering textbook series, and been the basis of a debate for a couple of hundred global institutional investors with an SMR advocate and analyst, Kirsty Gogan Alexander. What have I concluded about SMRs?
“Small modular reactors won’t achieve economies of manufacturing scale, won’t be faster to construct, forego efficiency of vertical scaling, won’t be cheaper, aren’t suitable for remote or brownfield coal sites, still face very large security costs, will still be costly and slow to decommission, and still require liability insurance caps. They don’t solve any of the problems that they purport to while intentionally choosing to be less efficient than they could be. They’ve existed since the 1950s and they aren’t any better now than they were then.”
The portion I was teasing apart with Flyvbjerg in discussion this week was that they “forego efficiency of vertical scaling.” What does that mean?
Well, thermal generation like coal or nuclear likes big boilers. It’s the inverse of liquid hydrogen storage having to be as close to balls and as big as possible, thermal efficiency is better in big globes because it minimizes surface area while maximizing volume. Want thermal efficiency? Scale up.
The nuclear industry learned that in the 1950s and 1960s when the US transplanted pressurized water reactors (PWRs) from nuclear-powered submarines and aircraft carriers to become commercial electrical generation facilities. The economics of subs and carriers is vastly different than the economics of electrical generation, and the very expensive power that had lots of benefits that the military liked was just far too expensive. So they scaled them up starting in the 1960s until they hovered around a GW in the 1970s and 1980s, which is mostly the standard globally. India was a bit behind that curve incidentally, as they were building 300 MW Candus until 2000 or so, but their new plants are GW-scale as well.
So the physics of thermal efficiency are important. So is modularity and manufacturability. There’s an optimizing curve in there that the SMR firms are trying to figure out, with some going tiny, and others like Gate’s Terrapower suddenly designing GW-scale plants, meaning that it’s not an SMR at all.
As I said to Flyvbjerg, the example of wind turbines and solar panels are very instructive in this regard. They are both well scaled to the requirements of the physics.
Starting with wind energy, wind turbines have been getting bigger, a lot bigger, with the passing of time. The Innwind project spent a long time and a lot of brain power seeing how they could overcome the engineering difficulties in scaling wind turbines to 20 MW each. Right now, onshore turbines are a lot smaller than that, averaging about 2.6 MW in nameplate capacity not because that’s the best size for generation, but because we can’t transport the masts, nacelles, and blades for bigger turbines to sites. The onshore limitation is purely logistics of transportation.
Offshore, most of those limits are removed because they can manufacture the turbines beside the ocean, put them on massive ships, and sail them directly to the wind farm location without worrying about little things like bridges or sharp corners along the route.
As a result, in both cases wind turbines have been scaled up to the maximize size that they can be due to logistical or engineering constraints. Longer blades, further off the ground are optimized for the physics and economics of the form of generation. And as discussed with Flyvbjerg, because hundreds or thousands of identical wind turbines are installed with four major components assembled on site, massive automation is performed to speed the process.
Solar panels operate differently from a physics perspective. The solar panel is actually made up of a bunch of cells. Those cells are actually more efficient when they are smaller due to the physics, but they are assembled in the factory into larger panels that are conveniently sized to put in shipping containers and be handled by one or two unskilled laborers.
Both of these examples are instructive. Wind generation has run up against engineering and logistics constraints, but solar clearly hasn’t. Among other things, solar panels have had a leg up from the massive containerization that’s spread through marine and ground shipping since the 1950s, while custom marine and ground vehicles have to be built for wind turbines.
But in both cases, wind and solar can be manufactured at scales which are very efficient for the physics of the form of generation. Wind turbines at 2.6 MW are already running at 95% of Betz Law. Solar panels are such cheap commodities that they don’t have to be more efficient than they are, although lots of money is going into trying to eke out a bit more because when you are delivering billions of something, a 1% efficiency gain turns into a lot of electricity.
SMRs are intentionally going back to the smaller scales of the 1950s and 1960s, scales that were tried and abandoned because they weren’t economical at that scale even with standardized reactors from subs and carriers, without any safety regulation to speak of, with no Fukushimas or Chernobyls in the rearview mirror, and with a national strategy that focused on nuclear energy, aligned with the US’ nuclear weapons program.
And note that SMRs don’t do a thing for the absolute worst case of projects in Flyvbjerg’s data set — nuclear waste storage facilities.
Will the SMR community or any firm be able to find an optimized point on the physics vs modularity curve? I don’t think so. Meanwhile wind and solar have very few fat-tailed risks in construction, are manufactured globally by the hundreds of thousands or tens of millions annually, have entire global supply chains, have master builders and skilled workers, are dirt cheap, and are completely reliable on well-managed grids at massive penetrations. I don’t think the nuclear industry has a hope of catching up.
Have a tip for CleanTechnica, want to advertise, or want to suggest a guest for our CleanTech Talk podcast? Contact us here.