This is a follow-up up on my previous post on renewable energy.
Aside: Apparently China is planning to build tons of new nuclear — does Xi read my blog?
Nota bene: I think coal is bad and we need to get rid of it. I also think many in the global north are unrealistic about how easy that will be for poor countries.
I had a fun argument with a colleague recently, about the outcome of COP26. Many (I mean, many elite climate-concerned COP-following types) in the wealthy world feel that India et al. are not doing enough; they should stop building new coal and go all-in on solar. Many in the developing world feel that the wealthy world should put their wealth where their mouth is and pay for it.
My simple point is that, as I argued in my previous post, solar energy could achieve an LCOE (levelised cost of energy) of $0, but it still needs storage, or interconnects, or something, if it’s going to let people watch cricket at 7PM. And right now, that something is more expensive than coal. And expecting Modi, or Ramaphosa, or whoever else, to bet their countries’ economies (and their citizens’ ability to keep food cold and run businesses) on untested strategies seems unfair .
My opposition said that solar + batteries would soon be cheaper, and in the long time it takes to build a coal plant, you could be way ahead with solar. So we made a bet on the following question;
In 2028, will coal still be cheaper than solar plus 32 hours of battery storage?
I meant coal or nuclear, but that’s not what made it onto the restaurant napkin. India was the context, but I don’t remember whether that was specified. 32 hours of storage is arbitrary, and probably more than is needed. We do need storage for 32 hours, but probably not covering every single GW of capacity. But the idea is that a day of no wind, and clouds all over India, might be followed by some big event. And load-shedding is a serious no-no. Again, there are additional solutions besides chemical batteries, but they are even less battle-tested.
And, as I went over in my previous post, India and the UK (representative wealthy country) have very different futures when it comes to energy. Chart data from here. The UK’s electricity consumption per capita is very high, and not growing much. This will change with more EVs and more heat pumps. But India is another story. You can’t tell because of the scale of the chart, but India is trying to hockey stick. And we want it to hockey stick: to get Indians to a vaguely UK quality of life, electricity consumption will have to 5x at least!
The chart below uses numbers from Lazard . Calculations are here. Solar and wind are much cheaper! They’re even sometimes cheaper than the marginal costs of nuclear and gas! But note, again, that this doesn’t include storage, or the fact that solar energy is only available when the sun shines.
LCOE is relatively simple, as you simply account for the megawatt-hours created and how much they cost. Storage is slightly more complicated, as it doesn’t actually generate energy , but simply time-shifts it from when it’s cheap or un-needed to when it’s expensive and needed.
Like generation technologies, batteries have a maximum capacity that specifies how much power they can output. Unlike generation technologies, the also have a maximum storage capacity before they’re depleted and have to be recharged. These two numbers are basically: how many lights they can turn on, and how long they can keep them on.
To compare coal with solar + batteries, let’s construct a scenario. A 2GW coal plant that we want to replace with solar. At a 50% capacity factor (it spends half its time switched off for maintenance and disruptions) it will generate around 8760 GWh of energy per year. The solar plant would have an installed capacity of 5GW and a capacity factor of 20% (most of the time it isn’t very sunny), generating the same amount of energy over the year. The costs for both of these, created simply by multiplying the values from the chart above, are shown below. Despite having a peak capacity 2.5 times higher, solar is still much cheaper!
Note: I have updated the following calculations thanks to some feedback! This doesn’t change my conclusion directionally, except to maybe narrow the margins slightly.
So let’s add batteries. We want something that can output at least
1 GW (the amount the 2 GW of coal can continuously deliver), and maintain it for 32 hours. So 1 GW capacity and 32 GWh of storage. (About the same as the total installed battery storage in 2021.)
This multiplier of 32 is far higher than typical grid-scale batteries, which are more likely to have around 5 hours. Based on this, I’m going to use purely the capacity costs from Lazard, but will have to do some juggling to come up with a reasonable figure for a 32 hour battery. The number is around 240 - 451 USD/kw-year for a 6 hour grid-connected battery. The chart below shows the following:
So… right now, the cheapest possible solar + batteries is around the same as the highest estimate for coal. Solar is probably on the more expensive end in India (not quite sunny enough), while coal is complicated as it comes out of the ground and is probably subsidised in all sorts of ways. But I would guess it falls to the lower end of this spectrum? So maybe today, in 2021, solar + batteries is around 3x more expensive. So that cost needs to drop by ~66% to become competitive .
The question is, will batteries drop in price by two-thirds by 2028? NREL has put together some projections that expect costs to drop 25% - 60% by 2030. Within reach of losing me the bet! These projections are often wrong, and sometimes spectacularly so. Some people  don’t think batteries will have the same dramatic price curves as solar, as it is limited by chemistry and the physical quantities of material that are needed to store a certain amount of charge. But these people are generally wrong.
So how sure am I in winning the bet… the stakes are high: the loser has to plant 100 trees, with the resulting carbon benefits accruing to the winner! And of course (labour obligations aside), I hope that I lose my bet. But do I think coal will be cheaper than solar+batteries (32 hours) in India in January 2028? I have no idea, but for posterity’s sake I’m going to put my credibility on the line and say
yes (coal will be cheaper), with
What if we include nuclear? Unlike coal, it might have a future of improved technology and learning curves… but probably not before 2028.
See you in six years!
 go back There is currently around 20 GW of installed battery capacity worldwide. A drop compared to the 2,000 GW of installed coal capacity. (But it’s growing fast!)
 go back Someone tell me if Lazard is suspect for some reason.
 go back Except don’t forget the First Law of Thermodynamics.
 go back I originally forgot to include the fact that coal’s 50% capacity factor means 2 GW can only provide 1 GW of continuous power.
 go back This is the levelised cost to provide this much capacity availability for a year, not the cost to install this many kW of battery capacity.
 go back The choice of logarithmic is I think supported to some degree by the final chart from NREL data, showing diminishing cost decreases as you get to longer duration batteries.
 go back A report on LDES (long duration energy storage) suggests costs must drop by 60% for LDES specifically to be competitive. I imagine LDES is cheaper than Li-Ion (or has the potential to be). Anyway this feels like confirmation that I’m in the right ballpark.