How the battery is separated from the connection it shares
NESO’s registers give one megawatt figure for each connection agreement. Where a battery shares its connection with a solar farm, a wind farm or a gas plant, that figure covers everything on the connection. Hornsea Power Station 3 is the clearest case: the register gives 3,000 MW, and the battery at Swardeston is 300 MW of it. Since 20 October 2026 the Atlas counts the battery alone.
What that does to the totals
The register adds up to 436,013 MW. Counting each battery alone gives 310,244 MW. The difference is the generation that shares hybrid connections. 103,179 MW of the battery total is our estimate, for connections where nothing public states the battery’s size, and the page says so on every one of them.
| How the battery figure was found | Projects | Register MW | Battery MW |
|---|---|---|---|
| Storage only, so the register figure is the battery | 887 | 192,637 | 192,637 |
| Stated by the developer, a planning document or a supplier, found by hand | 12 | 16,805 | 6,849 |
| From the capacity market register | 35 | 5,233 | 3,027 |
| From the government planning database | 44 | 9,139 | 4,551 |
| Our estimate | 610 | 212,199 | 103,179 |
| All projects | 1,588 | 436,013 | 310,244 |
Where a battery figure comes from, in order
A statement found by hand. Where a developer’s own page, a planning decision or a supplier states the battery for a particular connection, that is used first, with the source and the date it was read. Hornsea 3 is one: South Norfolk planning permission 2022/0867 and the switchgear supplier both give 300 MW and 600 MWh.
The capacity market and the planning database. The government’s renewable energy planning database lists batteries as their own entries, with an operator, a position and, since 2024, the reference of the generator each sits beside. The capacity market register lists every storage unit that has bid, with the connection capacity its owner declared. A queue row takes a figure from either only when the names agree and, in most cases, the owner agrees too. A word that is only the substation’s name never matches by itself, because several developers queue at the same substation: Daines, Monk Fryston, Staythorpe, Heysham and Melksham each have batteries by other owners. A battery already running is never matched to a row for new capacity. Where both sources give a figure the larger is counted, because a capacity market unit is often only the part of a site bid into one auction, and both are quoted.
Our estimate. Where nothing public states the battery, we use the share of the connection the battery takes on the hybrids where a source does state it. That share falls steeply with size. A 57 MW connection to a supergrid transformer’s tertiary winding is nearly all battery; above 400 MW the median is under half, and the spread is wide. So the estimate is read by size:
| Hybrid connections where a source states the battery | Cases | Median share | Middle half |
|---|---|---|---|
| Connections up to 60 MW | 40 | 96% | 66% to 100% |
| Connections of 60 to 150 MW | 15 | 50% | 24% to 100% |
| Connections of 150 to 400 MW | 10 | 49% | 29% to 76% |
| Connections above 400 MW | 23 | 47% | 14% to 100% |
Each estimated row carries the median and the likely range from the middle half, in its detail on the map and in the projects download, which now has battery columns beside the register figure.
What is wrong with it
The estimate rests on few cases, and they lean towards projects far enough along to have a planning application or a capacity market bid. A large early connection may be planned around a battery much larger or much smaller than the median, and the range on each row says how wide that is. A planning figure can be out of date, and for a refused or abandoned application it is only the size last applied for, which the row says. A capacity market figure can be part of a site. Every match we have read and found wrong is kept out by name, with the reason. If you know a battery figure we have wrong, tell us at [email protected] and send the source.
The front page, the country page and every finding that quotes the queue are computed from the same rows at every build, so they count the battery alone from 20 October 2026. Anything quoted from the site before that date counted hybrid connections whole.