> it's getting harder and harder to justify the efficiencies of centralization.
A lot of times they aren't even real to begin with.
People assume that economies of scale keep going up as long as scale keeps going up, but that's almost never true. They typically have diminishing returns or thresholds past which the unit cost stops going down. If you want to build solar panels you have to build a factory. If the factory can produce a million solar panels a year and you only want 10 solar panels you still have to build the entire factory. It's more efficient to build a million than 10.
But if people want a billion solar panels a year then you need a thousand factories, and one bigger factory isn't materially different than having two factories across the street from each other, so there's no real advantage to having them all operated by the same entity. Moreover, even if you only need 10 solar panels, you can get them from any of the thousand factories that each make a million a year. You're not losing the economies of scale by having many sellers and many buyers.
Meanwhile centralization often incurs additional costs. You already identified several, but another big one is land. Individual homeowners each have a roof or balcony wall that was otherwise going to have nothing on it. A centralized solar farm is more often going to have to pay for space.
Centralization is usually pushed by someone trying to monopolize something.
Well, there really are economies of scale involved in building solar farms. Accidental deaths per installed megawatt are orders of magnitude lower than rooftop solar. A single-axis solar tracker can rotate hundreds of square meters of solar panels. Washing dust off solar panels can increase their output by several percent, but is much more likely to happen if it's somebody's full-time job instead of a household chore, especially a household chore that puts you at risk of falling off a roof and dying. A FLIR image can identify failing solar cells so you can queue them for replacement or repair, and a lawsuit against a maker of faulty solar panels is much more feasible if the potential damages are €60 million rather than €600. Etc.
Also, clouds are less of a problem for a transmission grid with distributed solar farms than for an individual household with its own autarkic solar power system.
Even for solar energy, land is not a big cost, financially speaking. Morally and environmentally, it may be (it's arguable—solar farms don't have to devastate the ecosystem the way strip mining and oil spills do), but not financially.
> Accidental deaths per installed megawatt are orders of magnitude lower than rooftop solar.
This is 100-150 people a year. It's not even clear that this is more than the number of people who would die in car accidents on their way to work at centralized solar farms etc.
> A single-axis solar tracker can rotate hundreds of square meters of solar panels.
Those also cost thousands of dollars and it's not clear that it's a significant savings over the units that rotate fewer panels but cost less money.
> Washing dust off solar panels can increase their output by several percent, but is much more likely to happen if it's somebody's full-time job instead of a household chore, especially a household chore that puts you at risk of falling off a roof and dying.
This is a cost rather than an efficiency. If you get home and see dust on your panels you grab the hose and spray them off from the ground without having to pay anyone. The solar farm has to pay salary and benefits.
> A FLIR image can identify failing solar cells so you can queue them for replacement or repair
This is an inefficiency again. The centralized farm is paying for space so they replace panels with degraded output. The homeowner leaves them to run, gets 10% of the expected instead of 0% and if they want more capacity they get more panels instead of doing work to identify and remove existing ones.
> a lawsuit against a maker of faulty solar panels is much more feasible if the potential damages are €60 million rather than €600.
Class action lawsuits are a thing.
> Even for solar energy, land is not a big cost, financially speaking. Morally and environmentally, it may be (it's arguable—solar farms don't have to devastate the ecosystem the way strip mining and oil spills do), but not financially.
Land cost is why they can't put the solar farm near where the users are, because that's where the land is expensive, so instead they put it in the middle of nowhere. But even that land isn't free, and then you have to eat even higher transmission costs.
I think the answer is that they are, because there aren't that many infrastructure deaths.
https://www.ncbi.nlm.nih.gov/books/NBK448087/ says that in the US, where the statistics are best, "electrical injuries cause approximately 1000 deaths annually. Of these, around 400 result from high-voltage electrical injuries, while lightning accounts for 50 to 300 deaths." That's 400 deaths per year from high-voltage transmission lines and substations, and from other high-voltage sources such as CRT televisions being repaired or ion-implantation voltage sources. https://en.wikipedia.org/wiki/Electricity_sector_of_the_Unit... says the USA's utility-scale electricity generation was 4230.723 TWh in 02022.
So that's ballpark 100 nanodeaths per megawatt hour from transmission wires and the like. Or 0.1 deaths per terawatt hour. By contrast, https://www.nextbigfuture.com/2008/03/deaths-per-twh-for-all... claims that rooftop solar claimed 0.44 deaths per terawatt hour at the time; possibly that has improved since then, but I doubt that it has changed that much. Brian Wang returned to the question in 02021 in https://www.nextbigfuture.com/2021/07/2020-fatalities-for-us... and estimated almost 1 death per terawatt hour.
So it seems clear that the infrastructural deaths are much lower than the deaths from falling off roofs.
Secondary question, is rooftop solar installation any different in danger than housing construction or other equivalent blue collar construction jobs that people will fill anyway?
There are definitely economics of scale. They are just not visible in Germany due to outdated building and policy practices.
> I'm not talking about prices, which may incorporate subsidies, permitting costs, taxes such as tariffs, transaction costs, lawsuits against non-performing building contractors, and market inefficiencies such as homeowners not having access to the zero-marginal-cost excess power that can be produced on sunny days for regulatory reasons.
Every single thing here is policy failure by German government.
A lot of times they aren't even real to begin with.
People assume that economies of scale keep going up as long as scale keeps going up, but that's almost never true. They typically have diminishing returns or thresholds past which the unit cost stops going down. If you want to build solar panels you have to build a factory. If the factory can produce a million solar panels a year and you only want 10 solar panels you still have to build the entire factory. It's more efficient to build a million than 10.
But if people want a billion solar panels a year then you need a thousand factories, and one bigger factory isn't materially different than having two factories across the street from each other, so there's no real advantage to having them all operated by the same entity. Moreover, even if you only need 10 solar panels, you can get them from any of the thousand factories that each make a million a year. You're not losing the economies of scale by having many sellers and many buyers.
Meanwhile centralization often incurs additional costs. You already identified several, but another big one is land. Individual homeowners each have a roof or balcony wall that was otherwise going to have nothing on it. A centralized solar farm is more often going to have to pay for space.
Centralization is usually pushed by someone trying to monopolize something.