Leveraging AI for solar concentrator power system

BB Gun

[H]ard|Gawd
Joined
Jan 14, 2004
Messages
1,551
Why nobody else has done this in the last 10 years I have no idea, but at least someone has done it now. Leveraging AI to get superhigh temps on a solar concentrator array - enough to economically split H from H2O for hydrogen fuel.

https://q13fox.com/2019/11/19/secre...ed-by-bill-gates-achieves-solar-breakthrough/

Heliogen uses computer vision software, automatic edge detection and other sophisticated technology to train a field of mirrors to reflect solar beams to one single spot.
...
Heliogen said it is generating so much heat that its technology could eventually be used to create clean hydrogen at scale. That carbon-free hydrogen could then be turned into a fuel for trucks and airplanes.

"If you can make hydrogen that's green, that's a gamechanger," said Gross. "Long term, we want to be the green hydrogen company."
 
"If you can make hydrogen that's green, that's a gamechanger," said Gross. "Long term, we want to be the green hydrogen company."

Green hydrogen... That will be easier to spot than the transparent stuff when it leaks.
 
But, then you have to hire a bunch of off-planet hydrogen pickers...probably Jatravartids. And they don't work cheap.
 
  • Like
Reactions: dgz
like this
Seriously though, the amount of area for the mirrors and expense to auto tilt ect. It’s much easier and prob more efficient to just “plant” solar cell arrays.
 
its the temperature that's the thing, any solar concentrator can make power, just need a few hundred degrees to make water boil. This makes hi-temp heat... temps so far have only been available via carbon means. Carbon "free" (once you build the damned thing) super high temps for hydrogen separation, smelting, etc.

BB
 
yeah yeah yeah. not a panacea, but we'll see how it shakes out. Another piece of the puzzle, not an all encompassing solution.
 
But wait - it’s cloudy this week so shut production down.

We don't have coal-fired power plants in Idaho because it's inefficient to ship coal that far. Same-same.

The important breakthrough for this tech is making cement.
 
We have a solar collector array in California that heats up molten salt and then uses that to heat up water to drive traditional steam driven turbines. The molten salt keeps enough heat to keep the turbines running overnight. I doubt there is much benefit in turning it into hydrogen over just producing electricity.
 
I doubt there is much benefit in turning it into hydrogen over just producing electricity.

There's a few good reasons:

1. Hydrogen is much easier to store until needed.
2. Hydrogen can be pipelined (and even piggybacked on liquid natural gas pipelines) without loss, unlike electricity.
3. Hydrogen vehicles refill quickly, like an internal combustion engine (and they can be an internal combustion engine).
4. A hydrogen-to-electricity plant can make purified water.
 
What is the breakthrough here exactly?



Like this shit already exists, and has for years. Bill spent a few billion to just add "AI" to the front of it.
Yeah I'm curious about that too, it takes AI to aim mirrors? The path of the sun is a known quantity, the focal point of the mirror you design is a known quantity, "AI" simply is running the equations of the Sun's motion through the sky to motors that may track a bit more effectively than motors of the past? Maybe some feedback circuitry to change the speed if the motor gears move a little too much/little?


But whatever, if it can do something other solar concentrators couldn't do in past, then it's good regardless of the marketing hype. I'm guessing the "AI" aspect of it just allows them to file for an over reaching patent so that no one else can do this without paying them royalties. Saving the planet... by maximizing profit from others.
 
Why nobody else has done this in the last 10 years I have no idea.

Because the equations were hard to solve in real-time, and also the image processing has gotten a whole lot better in the last ten years.

It's tough to accurately focus light, while accounting for the atmosphere. The hotter it is, the more the air can mess with your focus, and it's increasing temperature as you move closer to the focus.
 
Because the equations were hard to solve in real-time, and also the image processing has gotten a whole lot better in the last ten years.

It's tough to accurately focus light, while accounting for the atmosphere. The hotter it is, the more the air can mess with your focus, and it's increasing temperature as you move closer to the focus.
Basically this, it has been something people have been working on for a long time but failing to do because they were limited by the technology available. Only in recent years have micro controllers, motors, and IR cameras reached the levels of precision needed to make the adjustments in real time. Same with onboard processing and small scale computing, having the sensors on the mirrors themselves is great but if you have to send said signals some half a KM away to a service station to do the calculations to then reposition them adds too much of a delay so they need to be done right there on the mirrors themselves. AI can then analize the success rate of various micro controllers and their inputs instead so the AI can analize things after the fact instead of having the AI try to do it in real time, the AI simply needs to learn to modify the algorithms that control the mirrors movements not the movements themselves.

This is a most interesting use of AI for me, I have a number of coluges who are currently training systems to find optimizations in mechanical processing and HR processes in an effort to streamline overhead and minimize form submissions. RIO Tinto's work with AI to increase electrical efficiency with Aluminium smelting is some really interesting work, and was also only recently made possible due to advances in optics and micro controllers.
 
Because the equations were hard to solve in real-time, and also the image processing has gotten a whole lot better in the last ten years.

It's tough to accurately focus light, while accounting for the atmosphere. The hotter it is, the more the air can mess with your focus, and it's increasing temperature as you move closer to the focus.

We already have this, power stations getting heated up to 1,000C already exist. Controlling the mirrors based on the sun.. already exists lol.
 
We already have this, power stations getting heated up to 1,000C already exist. Controlling the mirrors based on the sun.. already exists lol.


We ALREADY HAVE UNOPTIMIZED systems out there for decades now. This is meant to significantly improve focus.

JUST BECAUSE IT LOOKS EASY TO DRAW LINE BETWEEN TWO POINTS ON YOUR GRAPH PAPER DOESN'T MEAN IT'S THAT EASY IN THE REAL WORLD. You have a thing like superheated atmosphere to deal with.

The atmosphere around your hot-as-hell target ALSO gets hot-as-hell, and starts gusting around really aggressively. And that gusting air also provides different effects on refracting the light the hotter it gets. So yes you can "guess by drawing in a straight line," but it's not optimal.

If you want to optimize you need real-time image processing to create a real-time map of the atmosphere.
 
Last edited:
We already have this, power stations getting heated up to 1,000C already exist. Controlling the mirrors based on the sun.. already exists lol.
Yes there are systems out there that get heated up to and beyond 1,000C, but they are huge, expensive, and are not known for turning a profit. If anything they are more known for closing for not being profitable. If the technology is going to proceed in a way that isn't heavily subsidized they need to improve efficiency to reduce costs or they will always remain more expensive than mehtods using Natural Gas.
 
Last edited:
naysayers, you look at it too simply.

The atmosphere does affect the focus, and what they are doing is new. If they are smart yes they DID patent it, to protect their IP and R&D investment.

Example of a similar tech and why it is needed, to help illustrate why this is actually a real thing and not "we can point mirrors already, duh!": https://en.wikipedia.org/wiki/Adaptive_optics

Clear pictures of distant galaxies by ground based telescopes. Even on a perfectly clear night, the atmosphere affects the focus of distant objects. You can see this with your eyes as a twinkling visible on stars. Astronomers use a laser aimed at the upper atmosphere and measure the distortions in the reflection to calculate (on the fly) the proper focus to get the sharpest pictures. The upper atmosphere will have multiple layers of air at different temperatures, densities, moisture content, all of which have varying impacts to the focus. Probably something similar being done with this new tech, and rather than measure the effects on a laser which is not likely possible during the day (sun's too bright) they use AI to instead measure effects/correct the focus with the local information. Same source of the problem (atmosphere) but different method to solve it.

The [H]ate is strong here but I don't think it is justified.
 
  • Like
Reactions: Verge
like this
naysayers, you look at it too simply.

The atmosphere does affect the focus, and what they are doing is new. If they are smart yes they DID patent it, to protect their IP and R&D investment.

Example of a similar tech and why it is needed, to help illustrate why this is actually a real thing and not "we can point mirrors already, duh!": https://en.wikipedia.org/wiki/Adaptive_optics

Clear pictures of distant galaxies by ground based telescopes. Even on a perfectly clear night, the atmosphere affects the focus of distant objects. You can see this with your eyes as a twinkling visible on stars. Astronomers use a laser aimed at the upper atmosphere and measure the distortions in the reflection to calculate (on the fly) the proper focus to get the sharpest pictures. The upper atmosphere will have multiple layers of air at different temperatures, densities, moisture content, all of which have varying impacts to the focus. Probably something similar being done with this new tech, and rather than measure the effects on a laser which is not likely possible during the day (sun's too bright) they use AI to instead measure effects/correct the focus with the local information. Same source of the problem (atmosphere) but different method to solve it.

The [H]ate is strong here but I don't think it is justified.

Thank you sir, it makes much more sense now. I was confused as to what they were doing by reading the news article.
 
There's a few good reasons:

1. Hydrogen is much easier to store until needed.
2. Hydrogen can be pipelined (and even piggybacked on liquid natural gas pipelines) without loss, unlike electricity.
3. Hydrogen vehicles refill quickly, like an internal combustion engine (and they can be an internal combustion engine).
4. A hydrogen-to-electricity plant can make purified water.

Hydrogen suffers from major energy loss due to compression levels necessary for viable storage and transportation. People seem to overwhelmingly prefer BEVs over hydrogen fuel cell vehicles (see Tesla vs Mirai in CA). Really the only benefit I see from this tech is the ability to do advanced industrial processes without the use of fuels, but I question the ability to scale this up for manufacturing processes.

Of course, this may have other uses elsewhere, but I really don't see this technology being necessary for electricity generation from the sun.
 
Concentrating solar power systems are not a new breakthrough in technology.

CSP systems are comprised of multiple subsystems, and the solar collection field (arrays) that reflect sunlight to towers (which I think this utilizes) is just one subsystem.

This “AI” (but why?) system can only increase efficiency of the solar field. But you also have thermal losses in the chemical reactions/processes, losses in transferring the Hot heat transfer fluid media, losses in storage, losses due to operations and parasitics, and losses in the conversion from heat to electricity in the power block. I guess you can use AI for the novelty I guess.
 
Any energy production system has "losses". They've improved efficiency in one aspect of the overall process, probably a good thing. And when it comes to losses, the sun isn't going out anytime soon. And if they can collect and store as heat, enough energy to run the plant for 3 or 4 days in an 8 hour period, this becomes a much more viable and useful electricity generation method. Every bit of increased efficiency improves how long it can provide power and/or how quickly it "charges" or stores the energy for use over the next few days.
 
Back
Top