techno-economic analysis

There is a genre of engineering paper called "techno-economic analysis". Generally, they involve:

Listing multiple related designs for accomplishing something. These are mostly selected from previous literature, existing objects, patents, and information from companies. Sometimes novel variations are analyzed.
Finding key numbers for properties/costs/etc. People look at previous literature, markets and listed prices, datasheets, etc.
Optimization of various parameters, generally using specialized software. Sometimes the software is open-source, and sometimes (eg Aspen Plus) it's expensive.

Obviously such analysis can only be as good as its inputs. Ways that a techno-economic analysis can be bad include:

Analyzing something that doesn't matter.
Not considering the best relevant designs.
Using incorrect prices, eg: analysis concluding that a renewable chemical process is economically competitive relative to public prices for low-volume chemicals when real production costs used internally are much lower.
Using incorrect numbers from earlier literature, or from startups lying about something.
Not considering incompatible choices, eg: using a different fluid/metal combination that would give fast corrosion.

But while many analyses have such problems, I've still read a lot of them and found them very useful. The authors collect lots of key numbers for you from disparate sources, which makes them useful for the same reasons survey papers are. I understand other fields well enough to evaluate the assumptions and quality of designs used in techno-economic analysis papers, so I can pick out good ones and ground my intuitions for net costs.

Also, techno-economic analysis as a field has developed good norms, which I think is partly due to separation of analysis from innovation. If your analysis of a new proposed process shows that it's very expensive, that's fine; if your analysis is good, the high costs aren't your fault and it's still publishable. Meanwhile, if you look at university press releases (especially MIT ones) they'll often say: [trivial variation of earlier work] could lead to [technically possible but completely impractical application]!!! Computer security researchers have decent norms about designing secure systems, US chemical plant and aircraft designers have decent norms about designing safe systems, and techno-economic analysis has decent norms about estimating costs realistically.

about power generation

Electricity generation is foundational to modern civilization; costs are large but consumer surplus is much larger. Hedge funds may make more profit than power plants, but I know which is more important to civilization. However, I'm bemused by the people who think that cheaper electricity is the main thing holding back civilization. Look, California charges consumers literally 10x the production cost of electricity, but its economy has done OK, and its problems have different causes. In Germany, >half of electricity prices in 2021 were taxes. If the cost of electricity was really that important today, there are easier ways to bring it down than new generation technologies.

Solar-thermal power isn't currently very important, but there are some reasons I picked it as a topic:

Some people here are apparently interested in power generation.
It's renewable energy, related to global warming.
Large improvements from current installations seem possible, which is more fun than micro-optimization of gas turbine efficiency, and it's currently far enough from viability that I don't have to worry about saying something immediately valuable.
It's not potentially dangerous like military UAVs or bioweapons or some AI stuff.
I think they look cool.

solar-thermal analysis

Competent estimates for the cost of solar-thermal power are typically around $0.11/kWh. That's more expensive than US natural gas (~$0.04) and PV solar or wind in the US (~$0.03).

Things are actually worse than that, because such analysis usually assumes a sunny location, but a lot of power demand is in Europe and the US northeast. If you check a solar irradiance map, those aren't the sunniest places, especially in winter. Plus, clouds are worse for concentrated solar than for solar panels.

Yes, you can run a HVDC cable from Morocco to Europe, and people are actually doing that, but it's more expensive than burning LNG from the USA.

Also, the only reason for using solar-thermal power instead of solar panels is that storing heat is cheap, so you can use it to balance out renewables. But existing solar-thermal designs use steam, and using steam turbines intermittently is impractical.

OK, so if steam is out, then what? Here's a recent open-access techno-economic analysis (hereafter "Linares") of power tower type solar-thermal plants that use CO2 recompression cycles. There's a competent and concise example of a techno-economic analysis, so you can take a look and see what they're like.

Any time you have molten salt in heat exchangers, you have to consider corrosion, and chloride eutectics are generally worse than nitrates. So, they specified (nickel-based) Inconel 625 for heat exchangers, which is reasonable. But you have to keep the salt away from water and air, because oxychlorides are more corrosive. Corrosion is an issue for "solar salt" too, requiring (IIRC) stainless steel and (again) keeping it away from air/water.

improvements

Some people assume solar-thermal is less efficient than PV solar, but that's wrong; Linares gets ~50% efficiency.

A lot of people assume mirrors are what makes concentrated solar power expensive. That's wrong; mirror supports and drives are more expensive than the actual mirrors, and Linares has the entire solar reflector field at only ~15% of the total cost. Still, cost improvements are possible. Linares assumed $145/m^2 but $100/m^2 is feasible. (But the SunShot goal of $50/m^2 probably isn't.) Note that a typical US house today is ~$2000/m^2 of floor. Early heliostats used open-loop controls, which required stable bases and careful calibration. The trend now is towards closed-loop control with cameras, PV panels to power drive systems, and wireless connections.

What's more expensive, then? Per Linares Fig 15, more than half the investment cost is heat exchangers. I remember fans of molten-salt thorium reactors (weird thing to be a fan of) saying that "some fancy alloy has adequate corrosion resistance so heat exchangers aren't a problem". When I say people like that are clowns, part of what I mean is that they should do a proper techno-economic analysis.

Clearly, the conversion from concentrated sunlight is more expensive than the heliostats. If you want to get large cost reductions from optimized designs, you have to take a different approach, and there's actually a very simple and obvious way to greatly reduce the cost of conversion from sunlight: eliminate it. A lot of electricity is used for lighting, but people tend to prefer sunlight, and heliostats can focus light on skylights or windows. A few buildings have actually done that, but it's not very common; people like being able to see through windows as well as getting light from them, so generally building width is decreased instead of reflecting light, but it's actually sort of practical to use heliostats for building lighting, even if you still need artificial lighting too. It would've been better back when incandescent lights were a thing.

While it's not exactly economical yet, there are some compressed air energy storage (CAES) installations being built now. Combining a water-compensated CAES system with solar-thermal power cuts out some intermediate conversions, so you get better efficiency and lower cost, but potentially with more electricity transmission requirements.

Solar receivers on the tower are somewhat expensive, but there are 2 ways to mitigate that.

A boiling fluid means that you don't have to worry about even heating; sodium metal is sometimes proposed for that.
If you have a fluid with black particles in it, and transparent tubes (eg fused quartz) you can use a "direct absorption solar receiver" which isn't limited by heat transfer through metal walls, allowing for higher power density.

Heliostat costs have come down by >2x, but it has nothing to do with the "learning curves" finance types like to point at, it's just a matter of how much time smart people spent thinking about them. Steam turbines are expensive, but maybe you use CO2 instead and have turbines 700x smaller, but then heat exchangers are too expensive, so maybe you use supercritical ethane instead for lower pressure, or a different thermodynamic cycle entirely, or a thermal energy storage system with less-corrosive stuff that allows for cheaper materials, or something. Well, cost estimates from historical data don't mean anything without context. Even when people aren't using fundamentally new designs or technology, the costs of large construction projects vary greatly. Predicting the future is always extrapolation, and historical data is only useful as grounding for parameters used for that extrapolation, but with no technical understanding you're walking blind, and MBAs are liable to trip on a rock.

Anyway, as I've said before, it's possible to make power-tower solar-thermal cheaply enough to sometimes be worthwhile in sunny locations. I could get into details of designs I like, but haven't I posted enough on my blog already?