Bill Gates seems to love to invest in things that aren’t going to make much of a difference to climate change but that are good for the fossil fuel industry. The latest is Heliogen, a company which uses machine learning to make solar ovens hotter and more reliable. The problems are rife with the technology and it’s not competitive except in niches that likely aren’t actually climate friendly. While I’m a big fan of both machine learning for clean technology and solar power, this instance proves that they don’t necessarily produce useful results.
Previously, I’ve gone deep and wide on Carbon Engineering, which from my analysis and its sole contract is only useful for putting on tapped-out oil fields to burn otherwise unmarketable natural gas in order to push CO2 underground to get typically 2-3 times as much CO2’s worth of unrecoverable oil above ground. That company received at least one early round of funding from Gates. He’s also funding experiments in solar geoengineering, something which is good to have in our back pockets, but if used will simply perpetuate the burning of fossil fuels.
Let’s start with the good stuff. Heliogen actually has improved the performance of concentrated solar power. It uses four cameras pointed down at the array of mirrors to assess the light scatter around their edges to keep them tightly focused on the hotspot over the course of the day. This is a good example of something machine learning makes possible today that wasn’t easily possible a few years ago. This goes into CleanTechnica’s list of interesting uses of neural nets for clean technologies. And I’m sure that the method is useful in actually useful technical spaces too.
And there’s no reason to doubt that it works. Constant hyperfocusing of an array of mirrors to create a parabolic concentrator using this approach would create higher heat. It’s the same thing as a kid finding the focal point with a magnifying glass to fry an ant or make a classmate squeal. The ability to do this automatically and constantly means more rapid startup and a lot less downtime for calibration annually.
But that’s where it stops being particularly exciting too. This technology can create a single very hot spot, one that’s a long way off the ground. And it can only create it when the sun is highest. While it will generate lower levels of heat in the morning or late afternoon, it will only be able to generate its maximum temperatures for a few hours around solar noon on cloudless days.
There are five problems:
- It’s a single, poorly scalable disk of high temperature
- Tight coupling of energy production to energy consumption
- Industrial facilities cast big shadows
- Very high heat doesn’t store or transmit easily
- We have solutions already that are much easier to integrate into industrial plants
The first two are chunky enough by themselves, so will be covered in this article. The remaining three will be covered in Part 2. Also in Part 2, we’ll define where the characteristics of the technology would actually be useful and identify what is likely to be its only potentially viable use case.
It’s a single, poorly scalable disk of high temperature
The released prototype image with no supporting additional material shows a tower with a concentrated point of high temperature roughly 55 feet or 17 meters off the ground. The mirror array is in the range of 16,000 square feet or 1,500 square meters meters to achieve about 1,800 degrees Fahrenheit or 1000 degrees Celsius. That point of high temperature from images appears to be in the range of a foot in diameter or about 30 centimeters across. So we have a single, relatively small spot of high temperature a few dozen feet off of the ground. The current disk is in the range of about 0.4 square feet or 0.03 square meters.
That means that Heliogen has achieved a very high temperatures but not sustained heat throughout a large volume. And it’s heat, not temperature, that’s the requirement for industrial scale processes.
Want to double the area of the disk and hence the amount of heat available? That requires doubling the area of mirrors as well. That would give you a disk of high temperature about a foot and half or 45 centimeters across. It would require roughly 32,000 square feet or 3,000 square meters of mirrors.
The Ivanpah concentrated solar power facility, for example, had a capacity just under 400 MW with an area of mirrors roughly 3,500 acres or 1,400 hectares. That’s 5.4 square miles or 14 square kilometers, an area about a quarter the size of Manhattan. Its annual production in 2018 was around 700 GWh. And let’s be clear, Heliogen isn’t magic. The amount of solar energy that it can concentrate hasn’t changed and isn’t higher than what Ivanpah achieved, the technology can just focus it on a smaller spot more consistently and with less calibration downtime. Where Ivanpah achieved about 85% of its design parameters in 2018, Heliogen is easily closer to 100% based on the technology it is employing. That’s useful, if this is useful technology at all.
Let’s do some comparisons. The company talks about a few use cases, so let’s look at examples of these pieces of equipment to get a sense of the challenge here. Let’s start with cement.
One of the things that mentioned in various places is calcination of limestone into quicklime. That does require a lot of heat and is responsible for about half of cement’s emissions, which have estimates of CO2e emissions ranging from 5% to 12% of global CO2 annually, depending on what you look at.
You have to bring large masses to high temperatures, not just one part of them. Modern lime kilns require about 6–8 million Btu per ton of quicklime produced. A Btu is the heat required to raise the temperature of one pound of water one degree Fahrenheit. Alternatively, that’s about 6.3 to 8.4 gigajoules. To put it into more familiar terms, that’s about 1.8 to 2.3 MWh per ton of quicklime.
Lime kilns for large scale production of quicklime are rotary drums up to 13.5 ft or 4 meters in diameter and 400 feet or 122 meters long. Ignition of natural gas occurs inside the body of the rotating drum with high heat flames roughly 3 times the length of the diameter of the interior of the kiln, so in the case of the bigger kilns, that’s a jet of flame roughly 40 feet or 12 meters in length and about half the diameter of the kiln in width. The volume of the flame is roughly 1,300 cubic feet or 38 cubic meters. The interior of the kiln is maintained in the range of 600 degrees F or 315 degrees C for the entire day. Product takes 1-4 hours to transit the drum. Kilns at this scale can produce about 450 tons per day, so that’s the equivalent of 810 to 1035 MWh of electricity required.
So Heliogen can create a 1 foot or 20 cm disk of very high temperature for several hours, but processing cement at industrial scales requires a jet of flame 7 feet across and 40 feet long running all day long.
Let’s look at the energy requirements for quicklime vs the Ivanpah example, which produced sufficient heat for this process regardless, about 550 degrees Celsius in the receiver. Ivanpah was running about 29% boiler efficiency turning heat into useful electricity. If this could be made to work, the heat would be used more directly and efficiently in calcination, so let’s double that to give this the benefit of the doubt. For round numbers, we’ll assume 400 MW capacity for Ivanpah operating over 8 hours, giving about 3,200 MWh of electricity. Doubling that for useful heat for the quicklime process gives about 6,400 MWh of heat.
That suggests that the Ivanpah facility of 5.4 square miles would be producing sufficient heat to produce about 6-8 times as much quicklime as a single plant that takes a tenth of an acre. Turning that around, a single lime kiln powered by Ivanpah technology would take up about 0.8 square miles, about 5,000 times as much space. Assuming higher efficiencies from calibration, maybe Heliogen’s approach would only take up 4,000 times as much space.
So if concentrated solar power was useful for cement, then Ivanpah’s technology produced more than high enough temperatures, so there’s no specific advantage for Heliogen there. And the area required to generate the necessary heat is vastly larger than today for scaled industrial processes.
There’s a big mismatch here, and that’s problem one. The same math applies to pretty much all the rest of the industrial processes that the company talks about. We’ll talk about steel and hydrogen in another section.
Tight vs loose coupling
So what about tight coupling? Well, when I write about the future of energy being electricity, one of the points I make is that loose coupling is key to successful technical innovations. What that means is that the more you can allow major components to operate independently, the better the overall solution is, all else being equal.
There’s nothing in Bill Gross’, Heliogen’s founder, background that suggests he would have any particular reason to have this insight, but it’s straight out of software engineering, and Bill Gates should know this one inside and out, even if object- and component-oriented design came after his hands-on days. And Heliogen is a software-enabled innovation. It undoubtedly has people on staff who internalized this principal a couple of decades ago just as I did.
What this means for energy generation and industrial processes is that requiring them to be put together is tight coupling, and severely constrains the solution.
Consider an alternative. Instead of using concentrated solar, imagine you have a grid of electricity powered by renewables with a bunch of storage spread around the continent, and the lime kiln uses electricity to power its operations. That’s loose coupling, and it’s very effective. You don’t care what creates the electrons, you just care that you have MWhs of them coming in through the wires to your tenth of an acre facility. Instead of buying or leasing 4,000 times the land for partial power through the day, you have your tenth of an acre facility with wires coming into it and quicklime flowing out 24/7 if you want to.
There’s another point here. Transporting tons of durable goods is expensive, but transporting electrons is cheap and efficient. Land costs for a 4,000x area cement plan eat into profitability a lot more than the cost of electricity would. After all, a MWh for industrial users can cost as little as $45 USD, as it does for major customers in BC. That covers a lot when the alternative is most of a square mile of land somewhere.
For the solar solution for cement to be viable, it has to be in the middle of nowhere for land prices not to significantly change economic costs and that means you are transporting a lot more physical material a lot further at a cost per ton mile which starts to get excessive.
Heliogen claims that roughly half of all industrial sites globally have sufficient room for their solution, but that’s a gross overstatement as far as I can tell. If they need a fairly small amount of high quality heat for a few hours a day, that might be true, but industrial processes need not just high temperatures but lots of heat. They are too tightly coupled for this to be useful for most existing sites and it doesn’t seem that reasonable for economically viable future sites compared to just using electricity, with a bit more on that later.
That’s problem number 2, tightly coupling the primary source of energy to the industrial process instead of loosely coupling them to allow lots more sources of energy to provide the facility as needed.
In Part 2 we’ll cover the other the remaining three issues, the characteristics of a use case where this would be useful and the sole use case I can identify.
- CleanTechnica Report: A case study into a technology that should be set aside until 2050, Carbon Engineering’s air-to-fuel fig leaf
- Electricity is the future for all energy