From Trust, yet verify

Germany experienced two Dunkelflautes at the end of last year (a Dunkelflaute is a period with not much sunshine and not much wind). Wanting to know how many consecutive days they lasted, I started to search for more information, but my searches came out inconclusive. The German Weather Service (Deutscher Wetterdienst) mentioned a Dunkelflaute between November 2 and 7 (meaning 6 days), but I found other (less reliable) sources that also included November 8 (so 7 days). The Dunkelflaute in December was even more elusive. I only found one (not reliable) source that mentioned December 10 → 13 (meaning 4 days).

That didn’t go well. The only common information is that one Dunkelflautes took place in the beginning of November 2024 and the other just before half December 2024. I then decided to go to the Agorameter Energiewende website and look for myself what solar and wind did in those two periods. At first glance, I found more than 7 days of low solar and wind electricity production in November and about 3 days in December (11 → 13).

More confused than ever, I then went for confirmation to the Elia website to look at the Belgian solar and wind data. Belgium and Germany are neighbor countries, so they should have similar weather patterns. Here, I also found more than 7 days in November and the same 3 days that I found in the Agorameter data.

So, how long did these two Dunkelflautes last?

Looking further, I came across an interesting LinkdIn article detailing a quick statistical analysis on German dunkelflautes of the last 10 years. The article also mentions the Dunkelflautes in November and December of 2024.

The November 2024 Dunkelflaute is rather detailed in the article in this graphic:

It finds 12 consecutive days, which seems more in line with what I saw on the Agorameter data and the Elia data. A Dunkelflaute day is defined in the article as the combined electricity production by solar and wind of less than 10% of the installed capacity (red line with the text “10% CapFac”). The article explores two alternative definitions: a narrow definition (the number of days with a total production less than 10% of maximum capacity) and a broad definition (all periods with a production of less than 10% of capacity). The former definition is I think the most used and that is the one that I was interested in.

The Dunkelflaute of December is not so detailed, however. It is just described as a “few Dunkelflaute days” in the week of December 9 to 15 and it includes this image of the electricity production of the first half of December 2024:

I then decided to repeat the analysis of the November Dunkelflaute, see if I can reproduce the November graph and then apply the same to the December data to find the number of consecutive days.

The production data for solar and wind was easily downloaded from the Agorameter website. The capacity of solar and wind in 2024 was however not that straight forward. The article linked to the energy-charts.info site, but I was not able to find the information that I wanted. That tended not to be necessarily a bad thing. The author found this part a shortcoming in his analysis because the data he got from that site is of the beginning of 2024 and the period at issue is at the end of the year. In that year the capacity undoubtedly increased, therefore also the 10% threshold.

This gives me the opportunity to find more recent solar and wind capacity data for Germany. That was luckily easily found on the website of the German Federal Network Service. This is the (preliminary) capacity data from the end of 2024:

• Solar: 99.3 GW (coming from 83.1 GW)
• Offshore wind: 9.2 GW (coming from 8.5 GW)
• Onshore wind: 63.5 GW (coming from 61.0 GW)

This totals to 172 GW (from 152 GW) with a maximum potential production of 4,128 GWh per day (from 3,662.4 GWh).

When I apply this for the first half of November with the capacity of the beginning of the year, then I indeed get the same result as the LinkedIn analysis with the same 12 consecutive days. When I do the same analysis, but with the capacity of the end of the year, then I find two extra consecutive days:

Now I am rather confident that I am on the right track, I can do the same for the December data. When I do the same with the capacity of the beginning of the year, then I can identify those “few days” with a Dunkelflaute as December 11 until 13. When I do the analysis with the capacity of the end of 2024, then I get 4 consecutive days (December 10 → 13).

So, if the preliminary data of the Federal Network Service is correct and this capacity was active at the beginning of November and using the 10% threshold definition, then we are looking at 14 consecutive days in November (that is a lot more than I anticipated) and 4 in December. If that capacity was not (all) active, then the Dunkelflautes lasted at least 12 consecutive days in the first half of November and 3 in the first half of December.