Comment by epistasis | Hacker News Reader

Comment by epistasis | Hacker News Reader

Parent Full thread

I've just run my own numbers. But looking at your image, something seems seriously seriously off on the solar side. I get roughly similar numbers for nuclear; e.g. 250 thousand metric tonnes concrete / TWh for an AP1000 and 5x that for Sizewell B, using concrete mass numbers from [0]. Averaging these two designs is really close to their 760 number for concrete.

For solar, let's take some average random numbers, 800 pounds for a 6kW array on a roof [1]. So starting just with the panels, a 25% capacity factor over 30 years will generate 394 MWh == 0.000394 TWh. 800 pounds is 0.000363 thousand metric tonnes. So converting to the units of thousand metric tonnes / TWh, we are at 900 thousand metric tonnes/TWh for the panels for solar. Add another 10% for inverters, double that for framing if you're installing in a field rather than on a house, and I don't know where they are coming up with their stuff. 2700 tonnes/GWh for glass alone is ludicrous.

I'm not sure where that report comes from, but it would be worth checking the sources. After a basic sniff test, I'm going to dismiss it. Maybe it's using weights from 1980 solar panels or something.

[0] https://web.archive.org/web/20110722031413/http://www.nuclea...

[1] https://news.energysage.com/average-solar-panel-size-weight/

Edit: and honestly, both weight / Wh and land / Wh are fairly meaningless numbers for both technologies. What really matters is cost and environmental impact, since we have ample land and ability to put weight on land.

A couple points:

1. These numbers are from a 2015 report that was probably prepared in 2014 and calculated using data from several years before the report was written. Solar tech has improved a lot in the last couple decades. 2. The graph is measured in tonnes per TWh, not thousand tonnes per TWh, so that might be part of the confusion. 2700 tonnes per TWh is high, but reasonable. My estimate would be closer to 1000-1500 tonnes/TWh. 3. You are correct that mass is a pretty useless metric. A more important metric would be energy payback time (how long before the power source produces enough energy to make a copy of itself, essentially). For modern solar this is measured in months, possibly up to a year or two in extreme cases. For nuclear, this is probably measured in days or weeks, since nuclear power plants produce so much energy. However, when looking at power plants with lifetimes measured in decades, the difference is insignificant. Both nuclear and solar are fantastically better than any fossil fuel.

For EROI, "months" for deeply decarbonized renewable systems (i.e. what matters) seems to be in stark contrast to the values from this November 2019 study [1]. Is there a good critique of this line of thinking that you're aware of?

[1] https://www.sciencedirect.com/science/article/pii/S2211467X1...

I tried smell testing it too. For glass, we find 70 tonnes for a 1 MW capacity unit [1]. Using slightly more normal 20 year lifetime, and I'll re-use 25% capacity factor (which is about the avg in the USA), this comes out to 1597 tonnes/TWh of glass, compared to 2700 on the plot I showed. Running some sensitivity analysis shows that if you use 15% capacity factor, then you get 2662, basically the number given. Capacity factor is very location dependent so I guess the answer for solar resources is also highly location dependent. For instance if you're choosing between nuclear and solar in Boston, these numbers are pretty reasonable. In Arizona, there might be a factor of 2-3 improvement in solar's resource usage.

In any case, best-case solar uses vastly more raw materials from Mother Earth than all-case nuclear.

Also, solar is usually first built where it will make the most money, so early fleet capacity factors are likely to be biased high. They will decrease as people build solar in less sunny places.

[1] https://www.irena.org/-/media/Files/IRENA/Agency/Publication...

Your original graph is heartily contradicted by the IRNEA numbers. I used glass as a baseline because the bad QBER numbers had it as a small component, but it has become the dominant component in these builds. The original table (vs. IRENA):

2.9x (vs. 0.8x) as much steel as glass

1.4x (vs. 0.67x) as much concrete as glass

0.62x (vs. 0.55x) as much "other" as glass

Panels have 25 year warranties, assuming they will be replaced before the warranty ends doesn't seem very realistic. Choosing the worst-case-for-solar deployment of Boston isn't a good comparison metric either.

So all in all, the numbers that you have linked here show that solar, in its heaviest form deployed in an open field, is within a factor of 5 compared to nuclear on the weight per amount of energy over a lifetime. And various nuclear designs easily vary by a factor of 5.

SMRs, the only hope for nuclear going forward, have been avoided for a long time because it was thought they would be less efficient on a W/kg scale. So nuclear's future designs will take more mass per kilogram.

In contrast, solar is decreasing its weight and efficiency all the time, and there is no shortage of sites for solar, so we will see solar's Wh/kg decrease as the technology develops.

This is all a meaningless metric, but even if it did have some sort of meaning, it's not much of a win for nuclear.

Reply on news.ycombinator.com