How we estimate what a battery earns trading
A battery earns from four places. Two of them are published for each site and we simply read them across. The third, buying power cheap and selling it dear, is not published by anyone in Great Britain for any individual site. This page is how we estimate that third one, and what is wrong with the estimate.
What is measured and what is modelled
| Input | Where it comes from | Status |
|---|---|---|
| Money from being paid to change output | Elexon settlement | measured |
| Money from being paid to stand ready | NESO auction results | measured |
| Megawatts already committed to standing ready, half hour by half hour | NESO auction results | measured |
| Wholesale price, every half hour | Elexon market index | measured |
| How many hours of storage the site has | capacity market de-rating class | measured |
| What it actually traded | nowhere | modelled |
The calculation
Most published estimates take a battery's size, take the daily price spread and multiply. That counts megawatts which were not available, because a megawatt sold into a reserve auction cannot also be cycling.
We know exactly how much each site had sold, half hour by half hour. So the estimate is built only on capacity that was genuinely free:
- For every half hour, subtract the megawatts already sold from the site's registered power. What is left is free.
- Fill the battery from the cheapest free half hours of the day, and empty it into the dearest.
- Charge 15 per cent of the energy to round trip losses.
- Take off a share for the fact that nobody knows tomorrow's prices in advance.
- Take off £5 per megawatt hour for wear on the cells, which standing ready does not incur.
Three things we know are wrong with it
One. The share we take off for imperfect foresight is not calibrated. We compared our figures against a published fleet index for the two months where a comparison was possible. February 2026 matched to within 1 per cent at a 60 per cent share. January 2026 came out a third too low at the same setting. Two points that disagree do not calibrate anything, and until they do this is the single largest uncertainty in the number. If you have a view on what a real desk captures, we want to hear it.
Two. Our model can buy after it sells. It picks the cheapest half hours and the dearest half hours of a day without insisting that the charging happens first. A real battery cannot do that. This makes the estimate too generous and we have not yet measured by how much.
Three. Neither the gross nor the net balancing figure is really revenue. A battery is paid to turn up, and it pays to turn down. Reading only the first overstates what it earned, because it ignores what the energy cost. Netting the second off understates it, because the energy bought that way is not a loss, it is cheap stock that gets sold in the wholesale market at a profit we cannot see. In several recent months the whole fleet's net balancing figure is negative, which would mean batteries lost money in a market they were choosing to participate in. They did not. We publish both figures and label them, and we are still working out how to say this in one number without lying.
Other things it does not account for
- One cycle a day. Some sites do more, some do fewer.
- Standing ready is treated as blocking that capacity completely. In practice some of it can be stacked, so this makes the estimate cautious.
- The market index price is a reference. A trader also works within the day and through imbalance, which can beat it or miss it.
- No allowance for outages, auxiliary load, or a site being unavailable.
- Wear is charged at a flat £5 per megawatt hour. Real figures run from about £2 to about £10 depending on the cells and how hard they are worked.
Tell us we are wrong
We would rather be corrected than quoted. If you run one of these assets, optimise one, lend against one, or simply think the arithmetic is off, say so and tell us why. We will publish the correction and say where it came from.
A contact address for this is being set up. Until it is, the fastest route is whoever gave you this link.