From Trust, yet verify

It is a long time now that I wanted to have a deeper look into the cost of systems based on intermittent power sources, but didn’t know where to begin. That is why I was very interested when I came across a cleantechnica article by Georg Nitsche who created a renewable electricity calculator in order to prove that a system based on 100% solar and wind in the US is affordable.

This first post will be a 10000-foot view of this cost model and will look into the base data that it is working with. I will go into more details of how the costs are calculated in a later post.

This renewable electricity calculator is a Excel file with five sheets:

• about: the author, the version number (1.0) and the license (SPDX MIT).
• dashboard: one can enter numerous custom values and also a bunch of end results of calculations. That was the reason why I wanted to dive into this specific spreadsheet. Judging on this sheet, it promised to be a very detailed and extensive calculation.
• instructions: gives somewhat more detail about those custom fields and result fields, but despite its name it was not that helpful.
• calculator: is the workhorse of the spreadsheet. This is where the base data (solar, wind and demand data) and all calculations are placed. As neat as the “dashboard” sheet is, as chaotic this sheet looks. There generally is a lack of labels (40+ fields have no label whatsoever) and some fields that had a label were pretty vague/confusing. Some fields have conflicting values compared to other values in the spreadsheet. My impression is that these were used for testing, but not subsequently removed from the spreadsheet. It is as if the spreadsheet was abandoned after testing. All this meant that it took me quite a while before I found my way in the calculator.
• sources and figures: because there is no documentation, the content of this sheet was really helpful in finding some of the assumptions made in the calculator.

Nitsche created this spreadsheet because he wanted to show that a 100 percent renewable electricity grid is not only possible, but also affordable. He previously wrote a post on the same website in which he explained that cost estimations of a solar and wind system were too high because they are based on too little installed capacity of solar and wind, combined with batteries to fill in the gaps. He states that this combo wouldn’t work and I have no problem agreeing with that. I looked into such scenario before (using Belgian data) and also found that an insane amount of batteries would be needed to bridge the production deficits over one year, which wouldn’t make economical sense whatsoever.

To bridge the gaps when solar and/or wind isn’t working, he proposes to use Power-to-Gas, more specifically Power-to-Methane. That is an interesting idea and I could see this working in certain situations. However, I guess that it will come with a cost because of the (high) conversion losses from power to methane and then back to power. Yet, the scenario that Nitsche proposes ends up with a production cost of only 8 US cents per kWh (this is the cost of production of electricity, not the price that the consumer will pay).

The backbone of the calculator is US solar, wind and demand data. Nitsche uses 39 years of hourly data (from 1980 until 2018), so it ends up being a hefty Excel spreadsheet of 32.2 MB.

Let’s start with the demand data in the “calculator” sheet. What does this data represent? A link to the demand data is provided in the “sources and figures” sheet. Luckily, there was some more information in the header of that file:

The data is hourly Continental US demand data prepared by Tyler H and Dave F.. It was taken from July 2, 2015 to Aug 31, 2019 and then a four years loop is created over the rest of the spreadsheet So, it doesn’t reflect changes in demand between 1980 and 2018, it is rather recent data (at least at the time when the authors of that paper wrote their article) that is repeated over the entire dataset.

Then over to the solar data. Here I encountered the first hurdle. Just like the demand data, the solar data also resides in the “calculator” sheet and the column label is just “solar” without mentioning a unit. Trying to figure out how the data behaves, I found that the average of the first day (January 1, 1980) is about 0.14 for solar and 0.2 for wind, but weirdly enough, when I take the average of the last day (December 31, 2018), it is 0.14 for solar and 0.5 for wind. So there is some minor increase over 39 years showing in the wind data, but the solar data seems to stay roughly the same. What on earth are these values?

The sheet “sources and figures” didn’t give much clarity this time. A link to the solar source data is provided and its label is “US Solar generation“. Following this link, I landed on a csv file on archive.org. Also here, no unit is provided and this time the data is described in the header of that file as “solar capacities“…

Aaaarggghhh

This raises more questions than answers. These values are unlike any US production or capacities I could find. If these really are solar production or solar capacities, then why is there no increase in those values between 1980 and 2018? The BP statistical review of world energy has US solar capacity data since 1996 and there was an increase from about 14 MW in 1996 to 51957 MW in 2018. That is a 3740 fold increase in solar capacity not to be found in that column…

None of this made sense at the time. I contemplated that also some kind of loop might be used here, similar to the demand data, but at first glance this didn’t seem to be the case.

Looking for the name “Lei D” that is mentioned in the csv header and “Dowling 2020” that is mentioned in the url and at the top of the “instructions” sheet, I found the Dowling 2020 et al paper and, yes, in the acknowledgment section there was a thanks to Lei Duan (Lei D.?) and David Farham (Dave F.?) for providing the solar, wind and demand data. Also, the Tyler R. that provided the demand data might be Tyler Ruggles, a co-author of this paper.

Okay, that seems to be the paper where the data originally comes from.

Under “Experimental procedures” in that paper, I found a description of the data and the values in the solar column are in fact not “solar generation” as stated in the spreadsheet and also not “solar capacities” as stated in the csv file, but solar capacity factors. The authors of the paper estimated the hourly US solar capacity factors between 1980 and 2018 based on weather data from that time and these estimates are the values in the “solar” column.

That is why the compilers get solar data from as early as 1980, when solar was not much of a thing. This is not actual generation data, but estimates of what solar panels would have done based on weather data.

That is the reason why the solar data doesn’t increase over the 39 years. These values will always be between 0 and 1.

This also means that, when multiplying the capacity factor with the modeled capacity, this should result in the estimated production of solar assuming that this capacity would have been available between 1980 and 2018. This is where I bumped into the second hurdle: the capacity factors are not only multiplied with the modeled capacity, but there is also a weird correction factor of 6.4895930732 for solar ‘and 15.580509251 for wind)…

WTF?!?!

There is no mention of a correction factor in the paper and I also couldn’t find a clue in the spreadsheet itself. Where does these correction factors come from?

After some playing around with the data, it seems a correction that Nische introduced in the spreadsheet. He started with the assumption that the average capacity factor of US solar is 20%, while the data from the paper shows an average capacity factor of 27%. Using these correction factors, he comes to the production that would have been there according to the average capacity factors that he assumed in the calculator.

Now I got the solar data figured out, the same is true for the wind data. Its values come from this csv file also compiled by Lei Duan and are also the estimated hourly capacity factors based on the weather data from 1980 until 2018.

It took a while to get this far due to a lack of documentation combined with missing units and incomplete or even incorrect labels. Nevertheless, I laud Nitsche for at least trying to make a cost model and to share it with everyone. It makes it possible to look into the assumptions that have to be made to come to 100 renewables grid. In next post, I will look into the Power-to-Gas part, which I think is a crucial part in the calculator and also the part that I want to explore a bit deeper.