A Different Kind of Solar Power Attracts Attention | Hacker News Reader

A Different Kind of Solar Power Attracts Attention

wsj.com

39 pointsprostoalex29 comments

> Another innovation suggests that artificial intelligence could make CSP more competitive. Heliogen Inc., a CSP company backed by Bill Gates, uses computer vision to track the sun and continually adjust the mirrors in a CSP plant to maximize the concentration of sunlight.

Is this a journalist not getting it? Why would we need computer vision to track the position of the sun?

> That takes some of the complexity out of building a CSP system, which lowers construction costs. In current systems, mirrors have to be painstakingly installed in the best possible position. Even then, the ground can shift and stop concentrating the sunlight as effectively, according to Bill Gross, Heliogen’s CEO and founder.

Ok, maybe it's more about the continuous optimization of the mirror positioning.

It’s hard to adjust one mirror at a time, because 1000 other mirrors are reflecting at the same spot. And you can’t put a camera on the target to look back at the mirrors, because it would get melted. And some errors are harmless, like aiming too high, but aiming too low can melt the tower.

And you have to do all the aiming with cheap linear actuators that aren’t very stiff, so wind affects everything.

I dunno about AI, but it is a problem that programmers have struggled with so it might be worth a try.

> And you can’t put a camera on the target to look back at the mirrors, because it would get melted.

So just put it behind a heavy metalized filter that reflects over 99% of the incoming energy? And suspend the whole assembly about 1 meter in front of the target on a metalic post so it can cool down in the air. Or rather, a number of such assemblies so you can maximize incident energy over the entire target.

Optimizing that does not sound like a deep learning problem, just a classic computer vision algorithm.

Maybe. Just remember that an extra 1% of efficiency is worth 1% of the billion-dollar project cost, so there’s pressure not to have sensors blocking even a tiny part of the view.

Or just put the camera somewhere else. Then point one mirror at a time at the camera (therefore, it only seex 1x sunlight) to do angle calibration, and then aim that mirror back at the proper target and calibrate the next.

If actuators to keep the mirrors oriented correctly really do blow the budget then this sounds like a dead end.

It's not black & white. You want to use the cheapest actuators you can get away with. If clever software lets you use less precise actuators, it's a big win.

You could have position feedback on the mirrors and periodically calibrate the transfer function of mirror position to focus point at night.

I bet the mirror will heat up during the day and your night calibration will be off.

Good point.

Given the existence of passive solar trackers that work mechanically, it's really hard to see how a complicated active tracking system could possible outcompete for a PV system.

However, the company in the article is involved in concentrated solar power, so perhaps the situation is different enough to warrant such a system. I can _sort of_ see how there could be a feedback loop where changing ambient temperatures and radiative heat from the tower might effect the reflective array in complicated ways that would make it hard to maintain a very tight focus point. And, if tightening the focus point of the array can double or triple the incident W/m^2, maybe that gets past some sort of pass/no-pass temperature threshold that allows for higher overall energy conversion....

> Why would we need computer vision to track the position of the sun?

jpl horizons can tell anyone with a web browser where in the sky the sun will appear at any point in the next... hmm. ~477 years, it seems.

so there's got to be something more going on.

The early designs were required to redirect the mirrors on cloudy days. Sudden full sun would damage the tower.

The largest CSP has half a million mirrors.

https://archive.ph/GR2Ni

Concentrated solar costs 3x as much as PV if you want to make electricity.

But sometimes what you need is heat, and concentrated solar does it fine. For some reason not elucidated, nobody seems to be using it for industrial process heat, e.g. making cement.

http://www.brightsourceenergy.com/coalinga#.YnRFv-nMLUJ

Great, using solar to help put fossil carbon into the atmosphere.

That experiment ended in 2014 I think

I wonder how exactly one could use solar to make cement. The Portland cement process requires heating up ingredients to 1500 C. The concentrated solar systems as they actually exist do not get anywhere close to that. You could get this high temperature from concentrated solar, but I don’t know how you could effectively transfer the heat to the ingredients: I don’t think that there exists a practical pumping technology that can work with this hot working fluid, as this is literally hot enough to melt steel.

There are working fluids (molten salts, liquid metals, etc) that can work in this regime, typically with ceramic piping systems (usually ceramic lined nickel alloy steels) with all ceramic/graphite pumping systems. That said, it’s used in very niche applications and hard to maintain at the scale required for a CSP plant.

If you’re after heat, it would be much simpler to just use the steam. Use it as is, as we have for centuries now, or you could transfer it to a higher thermal capacity oil, and if you want some energy, divert it to a steam turbine.

For cement you could presumably use the heat as a preheater, or use the energy for an arc furnace.

https://www.cemex.com/-/cemex-and-synhelion-achieve-breakthr...

first attempt on making solar clinker, from earlier this year

Say it together: "solar thermal powered superheated steam bottle rocket".

We build a big, black bottle with a bunch of nozzles on the end, and focus these mirrors on the nozzly-end of the big, black bottle. You can use these "solar tracking mirrors with AI" to account for the fact that your tower is now going up faster than it normally does.

And the largest ones in Ouarzazate (pronounced Where Za Zat) use at least 2000 af of water annually.

https://www.cemex.com/-/cemex-and-synhelion-achieve-breakthr...

    >February 2, 2022

    >For the first time ever, CEMEX and Synhelion successfully connected the clinker production process with the Synhelion solar receiver, producing solar clinker.
    >This revolutionary innovation is an initial step to develop fully solar-driven cement plants.

they are trying it out, hopefully it works out for the whole manufacturing process, at an economically feasible rate

---

some papers on solar heat:

https://www.sciencedirect.com/science/article/abs/pii/S03605...

https://www.irena.org/publications/2015/Jan/Solar-Heat-for-I...

I always seem to see that plant when flying over California/Nevada, it's really easy to spot.

I thought there was a problem with birds getting cooked in mid flight with it or something though?

How was it flying through the blinding cone? Genuinely curious

It catches your eye for sure

It's the kind of thing you notice even if you're in the aisle seat basically because it's just a bunch of mirrors.

There's one of these in South Australia that grows tomatoes from sun and sea water, its been running for a while https://en.wikipedia.org/wiki/Sundrop_Farms

It would be better to concentrate all of the light into a space mirror facing the Kremlin.

Reply on news.ycombinator.com