The Life Of Lithium

I wager differently. I see no reason why sodium-ion batteries couldn't leapfrog lithium, just based on simplicity and economy. I suspect the learning rate will be much quicker for sodium ion.

> Most car batteries don't get drained on a daily basis. It's more like once a week or two weeks for average drivers. Same with grid batteries.

That's only because grid batteries aren't really being used to offset peakers. They are basically arbitrage for when prices are very high, as lithium batteries can hold charge for a very long time and can be dispatched almost instantly. In terms of batteries being used in place of gas and coal stations, they will need to cycle deeper, but maybe after about 6-12 hours of capacity is reached, they won't have to.

That said, the graph I provided is global capacity. Right now, batteries are still a specialty item, but for broader applicability, they'll be needed in more diverse scenario, including where there could be large gaps in renewables output. But even anywhere, an emergency capacity of days would become necessary in order to divest fully from non-renewables as both baseload and peakers (e.g., when overcast, still, and very cold/hot temperatures, or powerlines failing etc.)

> As for the IEA, take their numbers with a grain of salt. [...] Generally, you can subtract at least a few years from most of their predictions.

My projections were based on historical data, and regressing exponential curves onto them. Not super scientific, but the exponential trends are very visible for solar, wind, and storage. The linear trend for energy consumption seems to be holding now for decades, but that could also change.

> Apparently we can look forward to over production and price drops next year already.

I've already seen the prices drop ~40% from August last year to January on prismatic LiFePo4. I bought twice as many cells for just a little extra. I expect to buy again next year and double my capacity for about the same as my first purchase. We'll see.

The reason you care is that these stats aren't just abstract quantities; you're trying to gauge what the impact of a specific intervention might be. Suppose you could magically flip a switch and transition the country to EVs (which I'm not arguing against, just potentially the fire data thing), and you based that decision on this data. You weren't sold before, but since the fire risk is so low you were willing to apply that intervention. Instead of low fire rates, you might see:

1. When EV/EV collisions happen instead of EV/ICE, the higher total energy from heavier vehicles makes short-circuits and fires more likely than your model predicted (this is different from most of the rest of my points since it assumes your model is actually correct and unbiased today in some meaningful way but would still fail to be very helpful).

2. The data was just wrong and thus had no bearing on the real world.

3. You have the same fire rate as shoddy manufacturers enter the mix.

4. You have low rates initially, trending toward the same rate we're at now as the cars degrade, except now those fires are incredibly dangerous and toxic.

5. A mandatory software update while you're driving down the road inadvertently triggers a short circuit, leading to unknown-unknowns potentially affecting large swathes of the population at once (unlikely probably, but you are exposed to new failure modes and don't have a ton of data about the rarer ones yet -- much like (1) this isn't actually a criticism of the stats themselves, just their interpretation).

6. When you get more people driving these things in rural, snowy, mountainous regions you get more collisions and fires.

7. Having more teenagers driving EVs substantially reduces any claimed fire hazard deltas.

And so on. Global stats (like fires per crash, per mile, per car, ...) are suggestive of the next place to look for more data when trying to make an informed decision, but when you're talking about something affecting hundreds of millions of people, they're the beginning of the conversation, not the end. That's doubly true when they obviously have some kind of glaring flaw (like the napkin-math observation that they're off by at least an order of magnitude, probably 2 in this case).

Separately, once that data is interpreted correctly, you'll probably find other important observations. E.g., if ICE and EV have the same fire rate excluding shoddy manufacturers, and you have more solid evidence backing that up, that gives you very easy followups. The conversation can shift to the severity of those fires, their mitigations, and whether they're worth the cost given the other benefits EVs have. Moreover, you learn that you can greatly reduce car fires by just having better engineering, so (carefully and thoughtfully) toss in a law or tax or something to try to encourage better behavior.

Similarly, if you find old vehicles are the root cause then you can encourage recycling or rust-proof coatings. That might even be a point in favor of EVs since the batteries don't last long, so they'll be recycled before they're likely to combust (again, assuming for simplicity that everything else stays constant, which a policy-maker hopefully would not).

Just the stats presented though, especially without evidence that the obvious confounders don't apply, would likely not be helpful in making an informed decision. You could do just as well, perhaps better, from a gut instinct, and the presence of those stats in isolation is (charitably) just to help provide to people the data we do have and encourage further discussion, or (uncharitably) to present decontextualized information in a way that biases the populace in an intended direction, despite the fact that it's useless for the stated goal.

Even for the same car it's not trivial to use these all in the same system. It might even make most sense to couple them on the AC side but that would mean a lot of small expensive inverters.