Reactions from ranked influencers
3 posts
People building fission SMR's, held educate my grug-level engineer brain: - The usual narrative is that achieving economies of scale by mass producing small reactor sizes assembly-line style will bring down the $/MWh for Fission - But, the BOM cost per MW goes down as total MW goes per reactor goes up, just thinking shielding is like R^2 and total power like R^3 (grug-level analysis but first order correct) - Then balance of plant, control systems, safety interlock, facilities, siting, permits, and so on are more fixed-costs in some sense for any site or scale linearly with number of reactor volumes - would think favor fewer large reactor sizes per-site So, at what multiple of production volume does a 50MW reactor become cheaper than a 1000MW reactor per MWh? If the plan is to place 20x50 MW reactors at a single site, is that cheaper than one 1000MW reactor? Projected growth in power demand likely means we need many 100s of new 1000MW size reactors So is the 20x50 MW reactor facility still cheaper if you do it 200 times, versus doing 200 x 1000 MW reactors? Is there some 'master plan' like - Phase One: Mass-produce 50-200 MW sized reactors Phase Two: Mass-produce 500 MW reactors Phase Three: ... move on to 1000 MW reactors and so on, to ride the economies of scale and operations learning curve? I would guess there is some breakpoint in size where you have to move from factory production to on-site construction, but equivalently, some production volume required before factory-made beats out large-capacity reactors in $/MWh Gold standard for a civil nuclear power program is France, and their smallest size is 900MWe, and the newest are 1,650MWe, 20x the size of an SMR Why do we think France is doing that if SMR's are The Way? Something as simple as "largest single-object that can be transported by road freight" - which was the design constraint for the Rolls Royce SMR, ~450 MWe @isaiah_p_taylor @MattLoszak @juliadewahl @DougBernauer @sameer_n_soni @CarllPerez
The cost of nuclear is very far apart from its material cost—ie it has a high idiot index. You correctly point out that lower power per reactor needs more physical stuff per MW, but in high idiot index regimes you need replication and learning curves to drive down cost. I do think that over time, SMRs will trend larger, but they’ll start at the size of maximum replicability.
@isaiah_p_taylor Digging into that a bit it looks like most of the idiot factor comes from construction and installation timelines rather than say the manufacturing cost of the RPV itself, so perhaps the SMR advantage in learning rate is more installation speed than manufacturing volumes.