Perhaps you’ve seen articles saying that Li-ion batteries are down to $132/kWh according to a BloombergNEF survey. That was heavily weighted towards subsidized Chinese prices for batteries used in Chinese vehicles. In 2020, the average price per kWh excluding China was over 2x their global average. But of course, journalists are lazy, so they just report the average.
In 2019, Tesla charged $265/kWh for large (utility-scale) lithium-ion batteries. That was almost at-cost, and doesn’t include construction, transformers, or interconnects, which add >$100/kWh. Their PowerWall systems are much more expensive, of course.
The cost per kWh also depends on battery life.
LiFePO4 batteries are often described as having a cycle life of 5000 cycles. How about degradation over time? LiFePO4 generally lasts for 5 to 20 years, depending on the temperature and charge state. (High temperature is worse.) That’s similar to the cycle life.
That’s not so bad, and Li-ion costs could come down somewhat. $200/kWh for LiFePO4 is plausible. Cycle life is OK, degradation over time is OK, so everything is fine, right? WRONG! A “SEI” layer forms in Li-ion batteries, which thickens over time, increasing resistance and reducing capacity but protecting the battery from degradation. Cycling the battery breaks up the SEI layers somewhat, which affects degradation over time.
But Li-ion battery degradation isn’t understood very well by most people, and investors and executives and government officials just look at the cycle life if they even think about degradation at all. Anyway, 5000 cycles is too optimistic, and I expect something more like 2000 cycles for LiFePO4 in a grid energy storage application.
Anyway, you can see this 2020 report from the DOE. They estimated $100 / kWh stored / year operated for LFP batteries. If you operate them every day, that would be $0.27/kWh.
Again, the problem is that sometimes it’s cloudy for a while in the winter, and sometimes the wind doesn’t blow for a while, so you need more than 1 day of storage. But if Li-ion batteries were $10/kWh you could have a month worth of battery storage and forget about nuclear. That’s not really physically possible, but some finance types did draw some lines down that far.
Are there other types of energy storage besides lithium batteries that are plausibly cheap enough (with near-term technological development) to cover the multiple days of storage case?
(Legitimately curious, I’m not very familiar with the topic.)
Yes, compressed natural gas in underground caverns is cheap enough for seasonal energy storage.
But of course, you meant “storage that can be efficiently filled using electricity”. That’s a difficult question. In theory, thermal energy storage using molten salt or hot sand could work, and maybe a sufficiently cheap flow battery chemistry is possible. In theory, much better water electrolysis and hydrogen fuel cells are possible; there just currently aren’t any plausible candidates for that.
But currently, even affordable 14-hour storage is rather challenging.
Please go into more detail the model you are using to arrive at “2000” cycles for grid storage applications. The data sheet testing usually has 3500-7000 cycles, usually defined as 20-100 percent SOC, usually at 1C charge/discharge rate, with failure at 80 percent of initial capacity.
Some LFP are better, some worse. Usually shallower cycles (say 40-60 percent) have nonlinearly more lifespan.
What explains the disparity? Heat? Calendar aging? The manufacturers all commit fraud?
Now, with 1 cycle a day, and if I assume that 4000 cycles is a real number, that’s 11 years. Calendar aging is significant for that service cycle.
I also am curious what you think of sodium batteries. With cheaper raw materials it seems logical the cells could drop a lot in cost, assuming almost fully automated production. I have heard “$40” a kWh and similar cycle life to LFP.
As I said, it’s the interaction of cycling with calendar aging. When you cycle only once a day, you’re cracking the SEI that built up over that time. For an example of a paper on that, maybe see this.
Ok, having examined the paper, I’m going to dismiss it as invalid evidence. It uses NMC cells. No grid scale batteries today use this cell chemistry. Tesla did offer NCA cells for a time, but has switched the LFP.
Do you have any evidence for LFP cells? Note that NMC cells have mayfly like lifespans of a nominal 1000-2000 cycles, depending.
$1000 for 5.12 kWh. Claimed lifespan 6000 cycles. There are many listings like this, a utility scale order would be able to find the higher quality suppliers using better quality cells and order a large number of them. So assuming 5.12 x 0.8 x 3500, that’s 14336 kWh cycled. 6.9 cents a cycle.
That’s for the 1 day storage and assuming good utilization. Price skyrockets for ‘long duration grid storage’, if we assume 1 cycle a year (winter storage), and the battery does fail after 20 years, then 102 kWh cycled, $9.74 a kWh.
What do you propose we use for this once/season duty cycle? Synthetic natural gas looks a lot more attractive when the alternative is $10 a kWh.
See other thread. Your evidence is unconvincing and you shouldn’t be convinced yourself. You essentially have no evidence for an effect that should be trivial to measure. You could plot degradation over about 1 year and support your theory or falsify it. It’s rather damning you cannot find such a plot to cite.
Plot must be for production LFP cells with daily cycling.
You are wrong about that, but here’s a paper testing LFP in realistic conditions over a shorter period of time. In the future, you should start by looking for papers yourself if you think one is unsatisfactory, and I will not be replying to your comments, and I really have no interest in interacting with you at all. Bye.
Do you have empirical evidence? Some grid scale batteries, especially of the “server rack” commodity style that use LFP, should have 5 years of life already and by your model about to fail. I would argue that such failures observed over the tens of thousands of them deployed in various grids would be strong direct evidence of an n cycle field lifespan. I do not see any data in this paper collected from field batteries, merely a model that may simply not be grounded.
Are grid operators assuming they have 15-20 year service lives or 5?
Yes, some people have needed to replace batteries in some large storage systems, and/or augment them with extra capacity to balance degradation. I haven’t seen any good data on this, because most operators have no reason to share it. Also, due to rapid growth, most of the volume of first replacements is still upcoming.
If I read the paper right, it refers to degradation similar to leaving the cells at 100 percent SOC. It should be immediately measurable and catastrophic, leading to complete storage failures. Do you not have any direct measurements?
It’s an extremely falsifiable thing, there should be monthly capacity loss and it should be obvious in 1 year the batteries are doomed.
Someone could buy an off the shelf LFP battery and cycle it daily and just prove this.
We generally provide a 10-year “no defect” and “energy retention” warranty with every current Powerwall and a 15-year “no defect” and “energy retention” warranty with every current Powerpack or Megapack system. Pursuant to these energy retention warranties, we guarantee that the energy capacity of the applicable product will be at least a specified percentage (within a range up to 80%) of its nameplate capacity during specified time periods, depending on the product, battery pack size and/or region of installation, and subject to specified use restrictions or kWh throughputs caps.
Assuming a daily megapack cycle, that’s 5475 cycles over the 15 year warranty period. This would line up well with the cited “6000 cycles” for some LFP cells used for solar energy storage. It also means a price per kWh of 265⁄5475 = 4.8 cents per kWh, not 27 cents.
In addition, the true expected degradation, assuming competent engineers at Tesla, is likely less than that, with real expected lifespans probably around 20 years in order for a 15 year warranty to be financially viable.
Do you have any data or cites that disprove this, and are they more credible than this source?
The Powerwalls are over-provisioned, which is part of why they cost $600+/kWh.
They don’t expect full daily charge-discharge cycles, and were willing to eat the cost of replacements for the fraction of people who did that, hoping for lower battery costs by that point.
The Powerwall 2 warranty is, if charging off anything except exclusively solar, limited to 37800 kWh on a 13.5 kWh battery that’s somewhat overprovisioned. Which is...about 2000 cycles. The warranty promises 70% of nominal capacity by that point. (By that point, resistance would also be quite a bit higher.) Even with those limits, I expect a decent number of replacements under that warranty.
How cheap do batteries need to be for economics to ignore everything else in favor of solar/wind?
Maybe limit to the 90 percent case, this would be for 90 percent of the Earth’s population. Extreme areas like Moscow and Anchorage can be left out.
First off, some reported costs are misleading.
Perhaps you’ve seen articles saying that Li-ion batteries are down to $132/kWh according to a BloombergNEF survey. That was heavily weighted towards subsidized Chinese prices for batteries used in Chinese vehicles. In 2020, the average price per kWh excluding China was over 2x their global average. But of course, journalists are lazy, so they just report the average.
In 2019, Tesla charged $265/kWh for large (utility-scale) lithium-ion batteries. That was almost at-cost, and doesn’t include construction, transformers, or interconnects, which add >$100/kWh. Their PowerWall systems are much more expensive, of course.
The cost per kWh also depends on battery life.
LiFePO4 batteries are often described as having a cycle life of 5000 cycles. How about degradation over time? LiFePO4 generally lasts for 5 to 20 years, depending on the temperature and charge state. (High temperature is worse.) That’s similar to the cycle life.
That’s not so bad, and Li-ion costs could come down somewhat. $200/kWh for LiFePO4 is plausible. Cycle life is OK, degradation over time is OK, so everything is fine, right? WRONG! A “SEI” layer forms in Li-ion batteries, which thickens over time, increasing resistance and reducing capacity but protecting the battery from degradation. Cycling the battery breaks up the SEI layers somewhat, which affects degradation over time.
But Li-ion battery degradation isn’t understood very well by most people, and investors and executives and government officials just look at the cycle life if they even think about degradation at all. Anyway, 5000 cycles is too optimistic, and I expect something more like 2000 cycles for LiFePO4 in a grid energy storage application.
Anyway, you can see this 2020 report from the DOE. They estimated $100 / kWh stored / year operated for LFP batteries. If you operate them every day, that would be $0.27/kWh.
Again, the problem is that sometimes it’s cloudy for a while in the winter, and sometimes the wind doesn’t blow for a while, so you need more than 1 day of storage. But if Li-ion batteries were $10/kWh you could have a month worth of battery storage and forget about nuclear. That’s not really physically possible, but some finance types did draw some lines down that far.
Are there other types of energy storage besides lithium batteries that are plausibly cheap enough (with near-term technological development) to cover the multiple days of storage case?
(Legitimately curious, I’m not very familiar with the topic.)
Yes, compressed natural gas in underground caverns is cheap enough for seasonal energy storage.
But of course, you meant “storage that can be efficiently filled using electricity”. That’s a difficult question. In theory, thermal energy storage using molten salt or hot sand could work, and maybe a sufficiently cheap flow battery chemistry is possible. In theory, much better water electrolysis and hydrogen fuel cells are possible; there just currently aren’t any plausible candidates for that.
But currently, even affordable 14-hour storage is rather challenging.
Please go into more detail the model you are using to arrive at “2000” cycles for grid storage applications. The data sheet testing usually has 3500-7000 cycles, usually defined as 20-100 percent SOC, usually at 1C charge/discharge rate, with failure at 80 percent of initial capacity.
Some LFP are better, some worse. Usually shallower cycles (say 40-60 percent) have nonlinearly more lifespan.
What explains the disparity? Heat? Calendar aging? The manufacturers all commit fraud?
Now, with 1 cycle a day, and if I assume that 4000 cycles is a real number, that’s 11 years. Calendar aging is significant for that service cycle.
I also am curious what you think of sodium batteries. With cheaper raw materials it seems logical the cells could drop a lot in cost, assuming almost fully automated production. I have heard “$40” a kWh and similar cycle life to LFP.
As I said, it’s the interaction of cycling with calendar aging. When you cycle only once a day, you’re cracking the SEI that built up over that time. For an example of a paper on that, maybe see this.
Ok, having examined the paper, I’m going to dismiss it as invalid evidence. It uses NMC cells. No grid scale batteries today use this cell chemistry. Tesla did offer NCA cells for a time, but has switched the LFP.
Do you have any evidence for LFP cells? Note that NMC cells have mayfly like lifespans of a nominal 1000-2000 cycles, depending.
Without evidence a rational person would dismiss your claim of ’2000 cycles’ and assume ~3500 (on the low end for LFP). In addition they would assume bulk prices for the batteries such as this. https://www.alibaba.com/product-detail/48V-51-2v-100ah-Lifepo4-Lithium_1600873092214.html?spm=a2700.7735675.0.0.19c0fbFofbFoZR&s=p
$1000 for 5.12 kWh. Claimed lifespan 6000 cycles. There are many listings like this, a utility scale order would be able to find the higher quality suppliers using better quality cells and order a large number of them. So assuming 5.12 x 0.8 x 3500, that’s 14336 kWh cycled. 6.9 cents a cycle.
That’s for the 1 day storage and assuming good utilization. Price skyrockets for ‘long duration grid storage’, if we assume 1 cycle a year (winter storage), and the battery does fail after 20 years, then 102 kWh cycled, $9.74 a kWh.
What do you propose we use for this once/season duty cycle? Synthetic natural gas looks a lot more attractive when the alternative is $10 a kWh.
The main mechanism discussed in that paper is about the anode SEI. And the anode is the same.
See other thread. Your evidence is unconvincing and you shouldn’t be convinced yourself. You essentially have no evidence for an effect that should be trivial to measure. You could plot degradation over about 1 year and support your theory or falsify it. It’s rather damning you cannot find such a plot to cite.
Plot must be for production LFP cells with daily cycling.
You know degradation of capacity and resistance isn’t linear, right? You’d need a 5+ year long test to get the complete data for that.
No I don’t know this and the curves I have seen are linear until below 70-80 percent capacity. Please cite evidence.
You are wrong about that, but here’s a paper testing LFP in realistic conditions over a shorter period of time. In the future, you should start by looking for papers yourself if you think one is unsatisfactory, and I will not be replying to your comments, and I really have no interest in interacting with you at all. Bye.
Do you have empirical evidence? Some grid scale batteries, especially of the “server rack” commodity style that use LFP, should have 5 years of life already and by your model about to fail. I would argue that such failures observed over the tens of thousands of them deployed in various grids would be strong direct evidence of an n cycle field lifespan. I do not see any data in this paper collected from field batteries, merely a model that may simply not be grounded.
Are grid operators assuming they have 15-20 year service lives or 5?
Yes, some people have needed to replace batteries in some large storage systems, and/or augment them with extra capacity to balance degradation. I haven’t seen any good data on this, because most operators have no reason to share it. Also, due to rapid growth, most of the volume of first replacements is still upcoming.
If I read the paper right, it refers to degradation similar to leaving the cells at 100 percent SOC. It should be immediately measurable and catastrophic, leading to complete storage failures. Do you not have any direct measurements?
It’s an extremely falsifiable thing, there should be monthly capacity loss and it should be obvious in 1 year the batteries are doomed.
Someone could buy an off the shelf LFP battery and cycle it daily and just prove this.
That is literally what the linked paper did as a basis for their modelling. But of course they used multiple cells.
Right, the actual batteries you can buy with Chinese EV grade lfp cells is what to test. There are many variables here.
I have found strong evidence falsifying your theory.
From Tesla’s 10-k filing : https://web.archive.org/web/20210117144859/https://ir.tesla.com/node/20456/html , which has criminal liability if it contains false information.
Energy Storage Systems
We generally provide a 10-year “no defect” and “energy retention” warranty with every current Powerwall and a 15-year “no defect” and “energy retention” warranty with every current Powerpack or Megapack system. Pursuant to these energy retention warranties, we guarantee that the energy capacity of the applicable product will be at least a specified percentage (within a range up to 80%) of its nameplate capacity during specified time periods, depending on the product, battery pack size and/or region of installation, and subject to specified use restrictions or kWh throughputs caps.
Assuming a daily megapack cycle, that’s 5475 cycles over the 15 year warranty period. This would line up well with the cited “6000 cycles” for some LFP cells used for solar energy storage. It also means a price per kWh of 265⁄5475 = 4.8 cents per kWh, not 27 cents.
In addition, the true expected degradation, assuming competent engineers at Tesla, is likely less than that, with real expected lifespans probably around 20 years in order for a 15 year warranty to be financially viable.
Do you have any data or cites that disprove this, and are they more credible than this source?
The Powerwalls are over-provisioned, which is part of why they cost $600+/kWh.
They don’t expect full daily charge-discharge cycles, and were willing to eat the cost of replacements for the fraction of people who did that, hoping for lower battery costs by that point.
The Powerwall 2 warranty is, if charging off anything except exclusively solar, limited to 37800 kWh on a 13.5 kWh battery that’s somewhat overprovisioned. Which is...about 2000 cycles. The warranty promises 70% of nominal capacity by that point. (By that point, resistance would also be quite a bit higher.) Even with those limits, I expect a decent number of replacements under that warranty.
The megapack is not a Powerwall. Powerwall uses nca cells. Megapack uses LFP. The cycles you calculated are correct for good nca cells.