Grid data · Europe
At its deepest, the 12 August eclipse cost Spain about 4.2 GW of solar output

Grid operators spent days bracing for the eclipse. It came, smaller and shorter than the biggest forecasts. Measured against clear days, Spain was down about 4.2 GW at the deepest point and Germany-Luxembourg about 1.9 GW. Most of it happened within a single hour of early evening.
- 4.2 GWSpain’s shortfall at 18:15 UTC, against clear days68% of expected output
- 1.9 GWGermany-Luxembourg’s shortfall at 17:45 UTCthe German operators expected up to 2 GW
- 5.8 GWfive zones combined at 18:00 UTCrange 4.1 to 6.5 GW
Spain's solar plants were producing 1,948 MW at 18:15 UTC on 12 August 2026. Take a clear day from the same week, move it to that date, and the fleet would have managed about 6,127 MW in that quarter-hour. So 4,179 MW went missing. That was 68% of expected output, and the eclipse's deepest cut into the Spanish system.
Germany-Luxembourg hit its low point half an hour ahead of Spain. At 17:45 UTC, output stood at 1,065 MW against a baseline of 2,997 MW, a shortfall of 1,932 MW.
Both figures come from actual generation published on the ENTSO-E Transparency Platform at 15-minute resolution, set against each country's own clear days between 2 and 22 August.
What the operators expected
Red Eléctrica put the maximum reduction on the Peninsula at approximately 5 GW, around 13% of peak demand on a typical Wednesday in August, and warned that cloud cover could change the final impact. The four German transmission system operators said feed-in would fall by up to 2 GW under a cloudless sky, against more than 110 GW of photovoltaic capacity installed in Germany. Europe-wide, ENTSO-E put the figure at up to 9.7 GW under clear-sky conditions.
Germany called it spot on. The afternoon of 12 August was clear there, they expected a 2 GW drop, and we measured 1.9 GW. Spain depends on the baseline: our lowest gives a peak loss of 3.2 GW and our highest 4.4 GW. Red Eléctrica said 5 GW, and that was too high on every baseline we tried.
How the baseline was built
At 18:15 UTC, Spanish output on clear days drops by about 290 MW a day through August, because sunset comes a little earlier each evening. That is enough to bias a simple average of the days before and after the eclipse.
Only clear days go into this baseline. A day qualifies when its output between 12:00 and 16:45 UTC adds up to at least 90% of an envelope made of the highest output recorded at each quarter-hour across the 21 days. Each qualifying day is then moved to 12 August along a straight line fitted to the trend at that quarter-hour, and the baseline is their median. The spread of those days sets the range.
Before first contact, 12 August itself reached 0.977 of the clear-sky envelope in Spain and 0.989 in Germany-Luxembourg, so the evening dip cannot be put down to cloud.
An eclipse at sunset
In Madrid the eclipse reached its maximum at 18:32 UTC, when the sun stood 7 degrees above the horizon, and that late hour did much to limit the loss. In Berlin, maximum came at 18:08 UTC, with 84.8% of the sun covered but the sun only 3 degrees up.
Solar plants were already on the steep part of their evening decline. The eclipse brought the evening decline forward instead of taking a bite out of a full afternoon.
| UTC, 12 August | Spain actual MW | Spain baseline MW | Spain shortfall MW | Germany-Luxembourg actual MW | Germany-Luxembourg baseline MW | Germany-Luxembourg shortfall MW |
|---|---|---|---|---|---|---|
| 17:30 | 12,876 | 13,834 | 958 | 3,413 | 4,771 | 1,358 |
| 17:45 | 9,308 | 11,260 | 1,952 | 1,065 | 2,997 | 1,932 |
| 18:00 | 4,968 | 8,745 | 3,777 | 126 | 1,653 | 1,527 |
| 18:15 | 1,948 | 6,127 | 4,179 | 85 | 730 | 645 |
| 18:30 | 1,040 | 3,892 | 2,852 | 73 | 227 | 154 |
Across the zones with usable data
Five bidding zones had both 15-minute data and a clear afternoon: Spain, Germany-Luxembourg, France, Austria and Czechia. Their combined shortfall peaked at 5,814 MW at 18:00 UTC. Taking the extremes of each zone's clear-day range, the combined figure could be as low as 4,053 MW or as high as 6,533 MW.
So the 5.8 GW and ENTSO-E's 9.7 GW measure different things. The European number was a clear-sky forecast for a wider group of countries.
Small in energy, visible in the control room
The eclipse made only a small dent in the energy total. Over its eclipse window Spain's solar shortfall came to 3,914 MWh, 1.3% of the 293.76 GWh of solar it generated that day. Germany-Luxembourg lost 1,482 MWh, 0.3% of 493.56 GWh, and 12 August turned out to be its highest solar day of the 21 examined.
The operational impact was still real. NESO issued a precautionary Electricity Margin Notice for the evening peak in Great Britain and cancelled it before real time.
Why forecast misses point to storage
The two day-ahead misses ran in opposite directions. Spain's forecast for 18:30 UTC came in 460 MW below outturn, so it slightly overestimated the eclipse. Germany-Luxembourg's forecast for 17:45 UTC was 877 MW above outturn, so it underestimated the dip. Either way, hundreds of megawatts arrived that the day-ahead schedule did not contain, and the operators had warned in advance that cloud could change the outcome.
Covering that difference in real time is a job for fast-responding resources, and grid-scale batteries can do it if they are built, qualified and used by the system operator. Sustainable Digital Assets Inc., issuer of SDA Token (SDA), lists battery energy storage systems among its proposed projects. They are proposals, not operating assets.
About the data
How to read the summary tables
All times are UTC on 12 August 2026 unless a column says otherwise. MW values are averages over the slot: 15 minutes for Spain, Germany-Luxembourg, France, Austria and Czechia, one hour for Portugal, Belgium and Switzerland. A shortfall is actual output minus the clear-day baseline, so it is negative when output was below the baseline.
-
Shortfall by zone and quarter-hour:
actual_mwis reported output,baseline_mwthe clear-day baseline,baseline_min_mwandbaseline_max_mwthe lowest and highest of the clear days behind it, anddeficit_mwthe shortfall;deficit_largest_mwanddeficit_smallest_mware the shortfall against the highest and the lowest of those days.deficit_vs_11_13_mwanddeficit_vs_ten_day_mware the same shortfall against two simpler baselines, the mean of 11 and 13 August and the mean of ten neighbouring days. - Peak shortfall and energy by zone: the deepest slot for each zone, its range across clear days, the share of expected output lost, the energy shortfall over the eclipse window and the event day's clearness.
-
Combined figure in the article: the 5,814 MW at 18:00 UTC
is the sum of
deficit_mwfor the five 15-minute zones in the quarter-hour table; its range adds theirdeficit_largest_mwanddeficit_smallest_mw. Hourly zones are not added to it. - Forecasts against actual output: the day-ahead and intraday forecasts for each slot, the gap to actual output, and the usual gap on clear days.
- Daily solar energy: solar energy per zone and day, 2 to 22 August, and whether the day counted as clear.
- Clear days: the days that formed each zone's baseline. A day counts as clear when its output between 12:00 and 16:45 UTC reaches at least 90% of the highest output recorded at each quarter-hour across the 21 days.
Summary tables
- Technical note as archived on Zenodo (PDF)
- Shortfall by zone and quarter-hour, behind Table 1 and Figure 1 (CSV)
- Peak shortfall and energy by zone, behind Figure 2 (CSV)
- Day-ahead and intraday forecasts against actual output (CSV)
- Daily solar energy and clear-day flag, 2–22 August (CSV)
- Clear days used for each zone’s baseline (CSV)
Get the full data
- Register on the ENTSO-E Transparency Platform and request access to its REST API, as the platform’s user guide describes.
- Download actual generation per production type for solar (A75, B16) and the day-ahead and intraday solar forecasts (A69) for 2 to 22 August 2026, for Spain, Germany-Luxembourg, France, Austria, Czechia, Portugal, Belgium, Switzerland, Italy, the Netherlands and Poland.
- For each zone, take the clear days listed in the clear-days table, move each to 12 August along a straight-line trend per quarter-hour, use their median as the baseline and subtract actual output. The notes above define each column.
Source: ENTSO-E Transparency Platform (transparency.entsoe.eu): actual generation per production type (solar, document A75) and day-ahead and intraday solar forecasts (A69), 2 to 22 August 2026, for the bidding zones in the tables. Modified: for each zone we built a clear-day baseline per quarter-hour, subtracted actual output and summarised the shortfall by zone, time slot and day; the tables contain those results and the actual values inside the eclipse window only. Retrieved 14 September 2026. Not endorsed by ENTSO-E.
Sources
- Red Eléctrica, press release on the eclipse, 6 August 2026
https://www.ree.es/es/sala-de-prensa/actualidad/nota-de-prensa/2026/08/eclipse-solar-medidas-preventivas-para-un-impacto-limitado-en-la-generacion-solar - TransnetBW for the four German transmission system operators, 7 August 2026
https://www.transnetbw.de/de/newsroom/pressemitteilungen/sonnenfinsternis-am-12-august-keine-auswirkungen-auf-die-stromversorgung-erwartet - ENTSO-E, transmission system operators prepare for the solar eclipse, 7 August 2026
https://www.entsoe.eu/news/2026/08/07/transmission-system-operators-prepare-for-solar-eclipse-to-ensure-secure-system-operation/ - Instituto Geográfico Nacional, total solar eclipse of 12 August 2026
https://astronomia.ign.es/eclipses-de-sol-y-luna/eclipse-total-sol-de-12-de-agosto-2026 - NASA, eclipse predictions (Besselian elements) for 12 August 2026
https://eclipse.gsfc.nasa.gov/SEsearch/SEdata.php?Ecl=+20260812 - NESO, Operational Transparency Forum, 19 August 2026
https://www.neso.energy/document/385841/download - ENTSO-E Transparency Platform
https://transparency.entsoe.eu/
Cite this article
Mikko Rautiainen (2026). At its deepest, the 12 August eclipse cost Spain about 4.2 GW of solar output. Sustainable Digital Assets Inc. https://sdafintech.com/articles/2026/august-eclipse-solar-output (archived copy: https://doi.org/10.5281/zenodo.24290240)
Archived copy with a DOI: https://doi.org/10.5281/zenodo.24290240