In my search of the way that a 100% renewable electricity grid could function, I came across another example of how such a grid could work. Remember, in the last series of posts, I explored a first example of a spreadsheet by Georg Nitsche, but his spreadsheet failed because its strategy was entirely based on averages, therefor not representative of an actual grid. The new example that I found didn’t seem to have this flaw. As far as I initially could see, it went sequential over the data. Yet, it came to some ridiculously low storage requirements.
The example was created by Nick Stokes as a reaction to
a spreadsheet by Ken Gregory showing that about 250 TWh of storage would be required in order for the USA-48 grid to function on only solar and wind. Stokes didn’t agree with the methodology and made a simple model to disprove the findings by Gregory.
Stokes disagreed because he considered the share of solar and wind in Gregory’s model too low and stated that the higher the share, the lower the storage requirement. This was also the criticism that Nitsche brought forward, that the share of solar and wind should be high enough. It was 1.56 times the demand in his spreadsheet, so an amount of about half the demand needed to be curtailed over the year.
Stokes made his calculations for multiples of the capacity of solar and wind in 2019 and calculated the following storage requirements:
| Multiplier capacity of 2019 |
Minimum storage needed (TWh) |
|---|---|
| 7.3 | 241 |
| 10 | 51 |
| 12 | 8 |
| 15 | 2.4 |
| 20 | 0.44 |
| 25 | 0.11 |
That 0.11 TWh at the bottom seems ridiculously low, so I wondered what strategy would reduce 250 TWh to a mere 0.11 TWh…
That strategy become clear from a comment in a WattsUpWithThat post, in which someone asks Stokes how he would cover the 250 TWh gap that was calculated by Gregory. This is the question and Stokes’ answer from that comment (bold emphasis mine):
“Gregory found that just to replace existing fossil fuel generation with renewables, taking a recent year not an extreme case, would take about 250 million MWh of storage. How would you cover that?”
OK, I looked into Gregory’s “breakthru” report to see why he gets such ridiculous numbers. The reason is that he replaces FF [Fossil Fuels, Ed.] with just enough W&S [Wind & Solar, Ed.] to cover year-average FF generation. This is not enough W&S to meet the seasonal peaks. So he says that enough storage must be provided to cover the seasonal demand fluctuation.
But of course seasonal fluctuation in demand is the same problem for FF and W&S. FF does it by providing enough generating capability to cover the peaks. Otherwise it would have to use storage too.
And so there would be enough W&S, not just to meet the average, but to meet the seasonal demand peak, just as with FF. Then the need to provide storage for seasonal demand variation goes away. Storage comes back to only have to cover W&S variability.
The strategy is to produce enough power to cover the seasonal peak and use storage to only handle the fluctuations in solar and wind production. I do agree that providing just the average of demand would not be sufficient in a 100% renewable grid and that the system needs to be overdimensioned to some degree, but how much overdimensioning are we talking about? In other words, how much overproduction would be generated by 25 times the 2019 capacity?
Luckily, Stokes had created an R-script to calculate the values in the table above. The calculation is rather simple. He first loaded in the Gregory data from 2019, cleaned it up, then went sequential over the dataset using the different multipliers, subtracting demand from production. When production was higher than demand (meaning a surplus), then nothing happens in his model. When production is lower than demand (meaning a deficit), then the required electricity is taken from storage. In the end, the lowest value in the storage column would be the required storage dimension.
That is a very simple, but effective way of calculating the storage requirement and only a very small script is needed. I envy such conciseness.
Something that I missed in the script was that it didn’t account for losses. The Gregory spreadsheet accounted for an 80% roundtrip efficiency, the Stokes script however charged and discharged the storage without any losses. He explained the reason why he didn’t account for the losses in the readme.txt file:
The R program follows the same logic [as the Gregory spreadsheet, Ed.], but doesn’t allow for losses. Since there is usually a large surplus during recharge, this doesn’t matter.
I initially didn’t understand this logic, but that became clear pretty soon what he meant by that when I started graphing that scenario. This is for example how the storage requirement would look like over the entire year 2019:
Only 14 hours of storage were required over the entire year (=8760 hours)! Mostly in the beginning of the year (January 1 from 21:00 – 23:00, January 13 from 20 – 22, January 14 from 18:00 – 20:00 and February 28 at 08:00), but the capacity was determined by the storage requirement in summer (August 4 from 06:00 – 09:00). The rest of the year, that battery didn’t need to provide backup electricity.
When graphing electricity production from solar and wind versus demand, it became crystal clear why hardly any storage is required in this scenario:
The system is considerably overdimensioned, the potential production is way higher than demand and that production dipped only slightly below demand during those four periods of the year. This is the detail of the first three dips in January:
Of course “there is usually a large surplus during recharge”. Potential electricity production in this scenario is 3.4 times the demand and most of the time there is plenty of electricity production during the recharging of the battery. Most of it would need to be curtailed anyway, so yes, the tiny amount of charging losses wouldn’t make any difference whatsoever.
As an aside, there are also discharging losses and storage losses that are not taken into account and these will increase battery size slightly, but not by that much in this scenario.
In closing, would this system work? In theory, yes. this scenario provides more than enough potential production throughout the year to meet demand and battery storage requirement is very low. I however doubt that it will be possible to determine the cost of such a system based on current costs. The dynamics of such a system are so different that I don’t think it is possible to transpose the current build cost of storage, solar and wind installations to such an overdimensioned system.
For example, solar installations and windmill owners buy these installations with the intention of making money by selling the electricity they produce when it is sunny and/or windy. If we get into a situation in which solar and wind need to be curtailed in about 70% of the time, then I doubt that prospective owners will have much interest in buying such installations.
The same with the battery. It is only used 0.16% of the time and this will need to pay off the investment. It might be quite expensive electricity coming from that battery.
The same problem as in the Nitsche spreadsheet also applies here. It is based on averaged USA-48 data, so not really suitable for the purpose of calculating the cost of a 100% renewable electricity system. This because it averages out production as well as demand, meaning assuming that any demand in USA-48 could be covered by any production anywhere else in USA-48, which is not exactly a realistic assumption.
This is an interesting thought experiment, but I have my doubt whether it is realistic.
via Trust, yet verify
January 31, 2024 at 03:36PM
Iowa Climate Science Education 