Hacker News Reader

I can’t imagine that’s how industries could actually run. I think they can benefit from cheaper costs to absorb excess electricity or reduce capacity slightly when it gets too expensive, but shutting down an aluminum smelting plant seems like a non-starter as they need to run constantly 24/7 and require long startup/shutdown times. Same goes for desal - you’re not going to economically produce sufficient amounts of water just trying to time the energy grid.

https://www.reddit.com/r/AskEngineers/comments/3g78nj/are_al...

I agree.

Nonetheless, some industry are willing to give it a test to see how feasible it is.

https://www.lightmetalage.com/news/industry-news/smelting/tr...

> “We have reinvented the electrolysis process for the production of aluminium. For the first time, we will be able to vary the energy supply during operation significantly. This will allow us to react to changes in the electricity supply, which will benefit the power supply to households in Essen,” says Philipp Schlüter, CEO of TRIMET. “As an aluminium producer, we are naturally an energy-intensive company. As such, however, we are also a valuable partner for the energy revolution.”

> The €36 million trial installation converted a total of 120 furnaces in hall one of its Essen plant, which will be able to consume either 25% more or 25% less energy for up to 48 hours. The energy requirement can also be reduced to zero for up to an hour, if necessary. This means up to 2,000 megawatt hours of electricity can be stored for use in the energy revolution.

-----------

It requires new smelters, new software, new engineers. Its all very expensive. But it might be worthwhile. Just don't go like... crazy... with the idea right? There's downsides, but the upsides are immense. Keep an eye on the costs and practicality, and give it a test. It might be worthwhile.

That's my outlook anyway. Maybe Aluminum specifically is run too close to 100% utilization to be useful (we do have a very predictable amount of Aluminum...), but maybe other industries have a more start/stop and feast/famine kind of setup that would benefit from off-hours automation for energy price optimizations and/or even grid-tie stabilization.

But there are _real_ Aluminum plants giving this idea a test. So this isn't "vaporware" or even "theoretical". This is a real test occurring today.

-------------

We have to ask ourselves: is a 2GW-hr battery the best thing to build? What if Aluminum-smelters could instead vary their load by +/- 25% with this new software/manufacturing mechanism? What costs more, the 2GW-hr battery or this software + manufacturing change to Aluminum plants?

In both cases, we get 2GW-hrs of "energy storage". EDIT: IE, varying your load each day is a form of energy storage that should be considered by members of society. It may very well be cheaper to achieve 2GW-hrs of virtual energy storage by upgrading industry, rather than trying to build impossibly huge amounts of Li-ion batteries.

Varying load by +/- 25% is neat but that doesn’t solve the “I need energy at night” problem for solar. You either need batteries or wind (but even wind is risky because you’re screwed during windless nights).

All these other ideas are worth exploring and I’m 100% in favor of them, but the only path to decarbonizing the grid is nuclear & I’m so frustrated by all the solar absolutists that think renewables can power 100% of the grid by 2050 just because solar panel costs are dropping. We’d get much closer to net 0 if we were building fission plants aggressively (cue all the solar proponents claiming it’s too expensive or unsafe even when it really isn’t and solar costs are intentionally calculated in a misleading fashion to ignore the buildout needed for energy storage or altering unrelated industries to make solar more attractive).

When the grid transition began in the 2000s the hope was that solar prices, storage and nuclear would all see massive cost declines that made a fast transition possible. This has occurred for solar PV. It’s reasonably plausible that it will also occur for storage, given the huge incentives and rapid increase in manufacturing efficiencies. It has not happened at all for nuclear and there’s currently nothing on the horizon that gives much hope this will change by the 2050/60 time frame. Even China, which is blasting away at nuclear, predicts it will only be 18% of their generation by 2060 (and that’s likely optimistic.)

Yeah, nuclear is hard because it was strangled in the crib for 50 years and it’s hard to make up that ground on time scales for net 0. It still remains the better bet for decarbonizing the grid and all the other neat ideas seem mostly like party tricks and nothing more. I think China in particular is interesting because there’s a good chance that they’re building enough capacity that their costs start to decrease & they become the world leader in fission plants. Specifically interested in what they’re doing with MSR designs because they should be much much cheaper than traditional HWR designs. But yes, it’s going to take us a minute to rebuild and restart our fission know-how as a species. It’s also telling that China is building a heck of a lot of fossil plants (natural gas & coal) in addition to lots of solar. That’s because they have a lot of electrification to do (they have a massive rural population & they also have much more transportation that needs to electrify to use grid power). Solar is a fine companion to handle daytime peak growth (which is why they’re reaching TW scale with solar) but that baseload growth remains a problem which is why they’re building so many fossil fuel and nuclear plants.

I don’t want to re-litigate the “fission didn’t fail economically, it was killed” perspective (which I find unconvincing: people have proven they’ll put up with much more dangerous things if they’re profitable.) But we need to replace the entire global fossil fuel industry in 20-30 years, and the only possible way to do that is with energy sources that can be mass-produced in a factory with all the scale and efficiency improvements and virtuous cost and production flywheel that entails. Solar and wind are on this trajectory. Storage is plausibly on an earlier stage of this trajectory. Nuclear might get onto this trajectory through modular reactors, but those don’t exist and are at the “maybe” research stage with huge questions still unanswered. By contrast, commercial battery deployment and production is economical and projected to hit 6.5TWh by 2030.

As a related datapoint on “party tricks”, China built more solar and wind in the first nine months of 2023 than all 26 nuclear reactors it has under construction. They’re likely to build even more next year. The need for baseload is a valid point, but the trajectories of renewables vs. non-modular nuclear are so different that it’s hard to see nuclear catching up enough to make a difference.

Given that China is building out more coal plants than all other countries combined, I think that’s telling about the efficacy of renewables to achieve net 0. They’re targeting 2030 for peak CO2 emissions because they’re trying to cheaply electrify as much of their population as possible as quickly as possible because the CO2 reduction efforts that will dominate 2050-2100 will impose significant economic growth challenges & China wants to make sure they’re in a more stable position to compete with the West.

Yes, China is building a huge amount of solar capacity, but AFAICT it’s all about absorbing peak growth as they add more consumers of electricity. Baseline growth remains a huge problem and China is solving that with a mix of nuclear and fossil fuels. I fully expect China will continue to ramp up nuclear as they get better at it.

And EVs are a huge iceberg problem for the electric grid as they currently don’t run off the grid and represents a massive amount of energy consumption that will be added to the grid. On top of that they consume a huge amount of battery production which means there’s not that much left over for grid scale batteries. We also have no existence proof of economical grid scale batteries at any meaningful scale. You also have to overbuild your solar by quite a bit so that you can charge the battery up when the sun is shining to time-shift that excess capacity into the night.

Solar is fine and it’s not a knock against it, but there’s simply no trajectory to reach net 0 CO2 emissions and you will end up with a hybrid grid. 18% nuclear would represent a huge CO2 reduction because all that capacity would otherwise be fossil fuels (because the alternative is not building batteries because that tech doesn’t exist yet). MSRs are a neat party trick but honestly I think China’s approach of SMRs is a far safer bet in terms of being a massive cost reduction, safer, and use waaay less water & are far less of a research project. My hunch is that they’re also standardizing their nuclear plant designs to keep costs in check. Keep in mind that France is 90% fossil fuel free in their grid even though nuclear only represents ~60% of energy produced.

I haven’t read anywhere that fission failed due to economics - it has always been price competitive with fossil fuels. It’s the regulatory burdens & concerns about safety (some valid, most not) that strangled its growth. Do you have any links suggesting nuclear construction is uneconomical?

The latest estimates are that China’s emissions will peak next year and then enter a structural decline [1]. The 2030 number is generally viewed to be obsolete now (that’s how fast things are moving.) This is even with the new coal capacity they’re building, which they’re structuring to be “idled” (through government payments not to generate.) They’re also building large amounts of solar capacity in the west and in deserts, with continent-scale HVDC interconnects.

Right now the (global) goal isn’t 100% renewable energy, and it won’t be even by 2030 or 2040: a grid that’s 70-80% renewable with 20-30% average fossil emissions will be a massive improvement and will keep us on track for decarbonization. The interesting question is whether battery storage will decrease in price fast enough to deliver that remaining fraction, or whether we’ll have to build the remaining 20% with fission by 2050/60. Batteries are already economical enough that we have 5GW deployed in CA for infra-day time shifting. It seems like an incredibly long bet to imagjne that after three more decades of technological and manufacturing improvements their cost won’t have dropped enough to make them a viable competitor for nuclear. For my own part I wish I could take the other side of that bet, since I think it would be a good one. But we’ll have to wait and see.

[1] https://www.carbonbrief.org/analysis-chinas-emissions-set-to...

> China built more solar and wind in the first nine months of 2023 than all 26 nuclear reactors it has under construction

Do you have a source for this? It seems wild, and I'd be happy to read more about it.

Data source is BloombergNEF which I don’t have access to, but you can verify the raw numbers. Here’s a chart someone made, which also includes their adjustments for capacity factor: https://www.dropbox.com/scl/fi/9ijkxxhuscrr3x22tewk2/File-De...

thanks!

> Varying load by +/- 25% is neat but that doesn’t solve the “I need energy at night” problem for solar.

*48 hours* is the part of that statement that solves the "need energy at night" problem. By allowing this industry to shift power-consumption forward, or backward, by 48-hours, you effectively build a battery. A very strange battery yes, but it means the plant can collect during peak solar (+25% power usage), and then cut power at night by -25%.

>we might be able to even do this for air-conditioning, washing machines, or car-charging (if you have a PHEV or EV).

No _might_ about it, this currently happens for A/C, and is just getting going for car charging (see EV Managed Charging as a category).

Also some functions of commercial/industrial facilities that can be ramped more easily, eg shut off half of the hallway lighting in an office building. I think I recall things like rock crushers being turned off, but I might be misremembering that.

Unfortunately, solar panels get less efficient when hot, which is when your AC demand increases.

Ideally you could offset this by having people who aren't well positioned for panels do energy storage: Buy in the morning, sell back in the afternoon. Practically I don't think the utility companies are going to cut them in on a big enough part of the profits to incentivize them to do so.

Sorry I stated my original point poorly, and it wasn't exactly what you were saying. My bad.

Demand response for A/C tends to look more like pre-cooling before the hottest hours, then reduce demand during the hottest hours. So it's not about shifting demand towards maximum production, but rather shifting demand away from the demand peak (aka, peak shaving). So it's similar but a little different.

>Practically I don't think the utility companies are going to cut them in on a big enough part of the profits to incentivize them to do so.

The commercial version of what I stated in the previous paragraph (ie, demand response/peak shaving) empirically already happens, and has non-trivial dollars attached to it. Again, maybe not exactly what you're talking about, but it rhymes.

> Unfortunately, solar panels get less efficient when hot,

Less efficient by what factor?

Based on my experience - about 12%. See https://shkspr.mobi/blog/2022/07/solar-efficiency-and-high-t...

That will depend heavily on local conditions, type of panel, etc.

Significant, but not devastating.

That's pretty cool! 10% was my educated guess. It's good to have some data points.

Would you say the same holds across months? Say, would cooler May have better production than hotter July? Or is July still ahead because of more sunshine hours, offsetting the hot inefficiencies?

It really depends on the weather. You can see my monthly generation at https://pvoutput.org/aggregate.jsp?id=83962&sid=74451&v=0&t=...

Sometimes May beats July, sometimes it doesn't.

If you're interested, I have released all my panels' data at https://gitlab.com/edent/solar-data/

Thanks. This is all very interesting, I love graphs like these.

I see the pvoutput page didn't pick up some days in May and June 2020. I'm curious why, since I can see them on your .csv dumps.

Probably because my Internet connection went down.

I'm in a hot climate, I find in general with my system that the length of the day dominates over the loss in efficiency so December (southern hemisphere) is always the peak generation for me. January we usually have storms and rain due to the humidity so that's not usually as good, but it seems to be pretty reliably due to that shading from the cloudy and rainy weather more than heat.

I don't remember the numbers but I did watch this YouTube video where a guy strapped on some fans to a solar panel and actually got a better yield fan usage included.

Now I'm wondering if it makes sense to cool them with water and reuse that heated water for washing and showers.

Around 0.3% less watt per degree C

It's usually given in the datasheet of the panel.

Energy costs vary not only over the day but per quarter hour in each hour, e.g. during the morning the last quarter is cheaper than the first for each hour. So you don't necessarily have to shut down the machines, maybe you can just move the most energy expensive phase into another quarter.

If you are brewing beer you have times where you have to heat it and times where it needs to sit at a given temperature. Try to optimize the process to heat during cheaper quarter hours.

In Australia we've moved all the way to five-minute (!) settlement in our electricity market. Used to be 30 minutes but they switched in 2021 to give better price signals for very fast response technology like battery banks, as well as the demand response.

Do you have a auction + continuous market as well? With same or with different granularity?

Europe has an day-ahead auction with 1h products and continuous trading with quarter hours as the smallest product (still varying from country to country).

The most egregious broken window fallacy I’ve heard is a vc claiming that cryptocurrency lets you “virtually” transfer energy from areas with low prices to areas with high prices (without power lines!).

You assume that it is all linear which it may not be. Osmosis at power P may produce X litres of water per time unit while osmosis at power 4P may produce only 2X litres per time unit. And thus be viable only when the electricity price is N times lower.

>”On an individual level, we might be able to even do this”

At Tendril, we did this in 2017. We called it Orchestrated Energy.