Do small modular reactors (SMR) hold real promise?

The economic analyses of the projected costs for small modular reactors (SMRs) appear to rely on two important assumptions: 1) that the plants will run at capacity factors of current nuclear plants (i.e., 70%-90%+) and 2) that enough will be built quickly enough to gain from “learning by doing” on scale as has occurred with solar, wind and battery technologies. The problem with these assumptions is that they require that SMRs crowd out other renewables with little impact on gas-fired generation.

To achieve low costs in nuclear power requires high capacity factors, that is the total electricity output relative to potential output. The Breakthrough Institute study, for example, assumes a capacity factor greater than 80% for SMRs. The problem is that the typical system load factor, that is the average load divided by the peak load, ranges from 50% to 60%. A generation capacity factor of 80% means that the plant is producing 20% more electricity than the system needs. It also means that other generation sources such as solar and wind will be pushed aside by this amount in the grid. Because the SMRs cannot ramp up and down to the same degree as load swings, not only daily but also seasonally, the system will still need load following fossil-fuel plants or storage. It is just the flip side of filling in for the intermittency of renewables.

To truly operate within the generation system in a manner that directly displaces fossil fuels, an SMR will have to operate at a 60% capacity factor or less. Accommodating renewables will lower that capacity factor further. Decreasing the capacity factor from 80% to 60% will increase the cost of an SMR by a third. This would increase the projected cost in the Breakthrough Institute report for 2050 from $41 per megawatt-hour to $55 per megawatt-hour. Renewables with storage are already beating this cost in 2022 and we don’t need to wait 30 years.

And the Breakthrough Institute study relies questionable assumptions about learning by doing in the industry. First, it assumes that conventional nuclear will experience a 5% learning benefit (i.e., costs will drop 5% for each doubling of capacity). In fact, the industry shows a negative learning rate--costs per kilowatt have been rising as more capacity is built. It is not clear how the SMR industry will reverse this trait. Second, the learning by doing effect in this industry is likely to be on a per plant rather than per megawatt or per turbine basis as has been the case with solar and turbines. The very small unit size for solar and turbine allows for off site factory production with highly repetitive assembly, whereas SMRs will require substantial on-site fabrication that will be site specific. SMR learning rates are more likely to follow those for building construction than other new energy technologies.

Finally, the report does not discuss the risk of catastrophic accidents. The probability of a significant accident is about 1 per 3,700 reactor operating years. Widespread deployment of SMRs will vastly increase the annual risk because that probability is independent of plant size. Building 1,000 SMRs could increase the risk to such a level that these accidents could be happening once every four years.

The Fukushima nuclear plant catastrophe is estimated to have cost $300 billion to $700 billion. The next one could cost in excess of $1 trillion. This risk adds a cost of $11 to $27 per megawatt-hours.

Adding these risk costs on top of the adjusted capacity factor, the cost ranges rises to $65 to $82 per megawatt-hour.

Advertisement

13 thoughts on “Do small modular reactors (SMR) hold real promise?

  1. Lee Kasten

    The solar/wind + storage costs are what we would call a Conditional Firm resource whereas nuclear is simply Firm. They behave the same most of the time but there are about 20 to 25 days a year where you have synoptic scale weather events which drive RE performance into the dirt. I don’t say this to pooh pooh RE. It’s just what the weather data says.

    I’ve done some basic modeling of high penetration RE grids. Your post made me ask what would happen in my model with and without nuclear. In both scenarios, I’m able to get residual loads down to around 125 to 135 TWh (about 3% of annual load). In the no nuclear grid the model wanted gas backup of roughly 370 GW – This is consistent with far more sophisticated models. Strangely, the model that has nuclear reduced the gas backup to 260 GW but this is less than the nuclear capacity (105 GW).

    The no nuclear model selects for 14 times more wind (878 GW) and 14 times more solar (370 GW). The nuclear model calls for 11 times more wind (690 GW) and 8 times more solar (211 GW). This means the nuclear capacity displaces 100 GW of gas, 158 GW of PV and 188 GW of wind. This is a substantial amount of capacity but you can channel check the result by pencilling in how much each resource would generate – i.e. 45 TWh from gas, 225 TWh from PV and 550 TWh for wind. This lines up nicely with the 800ish TWh we get from nuclear power.

    My modeling only included 200 TWh of new EV load and I didn’t model any thermal load or batteries. If you add all of this new flexible load it will greatly increase the amount of solar and wind you can install because it reduces your spill rates. IIRC you can get the gas back all the way down to 290 GW in the idealized models with scads of flexible load.

    FYI – This was not an econometric model. It’s just a sketch of the problem using idealized transmission and simple algorithms that scale and/or move existing generation.

    One of the things that’s interesting about these models is that they suggest you can get the residual load down very low (3%) but you still need to carry a lot of firm capacity. There may be ways to cleverly use things like V2G here or there to shave this down but the key problem is that the energy coming into the grid from poorly performing RE is much less than the energy going out.

    Finally, wind is considerably more variable than solar so hypothetically you could end up with 50 times more solar. In this type of a system I think the missing power problem (the gas backup) could be reduced – particularly with a small amount of V2G.

    Like

    Reply
    1. Richard McCann Post author

      The sequence of those 20-25 days is important. Most likely those days are distributed in groups of 3-4 days. When EVs reach market saturation, the low cost solution will be to have distributed stationary storage for the day-to-day needs and then use EVs as peak back up storage for a neighborhood. In California, the EV fleet should have up to 30 times the current peak electric demand in available storage, so EVs should be able to power a neighborhood for a month.

      This recent LBNL study shows that microgrids can be sufficiently reliable if loads are managed during those low RE periods. This study did not include EVs as a separate back up resource. In addition, it’s probably cheaper to use renewable gas fueled generators for those few days as an alternative as well. https://emp.lbl.gov/publications/evaluating-capabilities-behind-meter

      SMRs are supposed to operate in load following mode, which will increase the effective cost per kWh by 50%. And SMRs have to be adopted in large enough numbers to created learning by doing cost reductions. These two economic factors work against SMRs being widely deployed, especially if the grid is decentralized.

      Like

      Reply
      1. Lee Kasten

        3 to 4 days is correct but the distribution tends to be clumpy so you’ll have 3 to 4 days followed by a short break of 1 or 2 days and then another 3 to 4 days. This clumpiness isn’t 100% of the time but it’s common enough that you can see it in the data visualizations. At the total US grid scale the two challenging parts of the year are obviously winter and summer. The electrification of heating and transportation will make the winter much more challenging.

        Some places in California have mild weather so if anyone is going to do a microgrid it would be there. At the US scale the loads in the North East and Midwest dominate – during extreme weather episodes the gap between RE and load is very large in the modeling (85% unmet).

        Like

      2. Richard McCann Post author

        The question then will be are the RE outages diversified geographically so that increased regional transmission interconnections will get wind power to Boston when the wind is blowing in Iowa? Another is whether EVs trapped at home in the snow are ready multiday storage sources? This probably needs another level of modeling to dig deeper.

        Like

  2. Richard McCann Post author

    $60/MWh for advanced nuclear electricity is achievable, says GE Hitachi executive.

    Meanwhile, the costs of renewables and storage are falling. Utility-scale solar-plus-storage costs are about $45/MWh; wind power costs are $30/MWh; and stand-alone utility-scale solar costs are at $32/MWh, according to the Institute for Energy Economics and Financial Analysis. And that group has doubts that SMRs can be developed as cheaply as their backers claim.

    https://www.utilitydive.com/news/advanced-nuclear-ge-hitachi-mwh-nuscale-smr-small-modular-reactor/630154/

    Like

    Reply

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s