How accurate are the grid forecasts?
Every 30 minutes we save each official forecast as it is published, then check it against what happened. Nobody else publishes this, so here it is.
Live data, updated .
Carbon intensity: how clean it is
gCO₂ per kWh, past and next 24h
Cleanest window ahead 02:00 to 04:00, around 68 g/kWh.
- Day-ahead forecast
- Actual
- Forecast
the numbers
under about 100 g counts as low, over 200 as high
More on carbon intensity →Electricity demand vs forecast
GW, last 48 hours and the day ahead
- Day-ahead forecast
- Demand
- Forecast
the numbers
Day-ahead forecast typically off by about 1.7 GW (last 30 days)
dashed over the past is what was forecast a day before. this is grid-level demand, so rooftop solar is not counted
More on demand →Carbon intensity forecast, last 30 days
| Made this far ahead | Typical error | Bias | Half-hours |
|---|---|---|---|
| Under 2 hours | 15 g/kWh | −4 g/kWh | 464 |
| 2 to 6 hours | 23 g/kWh | −1 g/kWh | 891 |
| 6 to 12 hours | 30 g/kWh | +3 g/kWh | 1,177 |
| 12 to 24 hours | 33 g/kWh | +10 g/kWh | 1,656 |
| 1 to 2 days | 31 g/kWh | −1 g/kWh | 2,168 |
Forecasts from the Carbon Intensity API, compared with its own estimate of actual intensity.
Demand forecast, last 30 days
| Made this far ahead | Typical error | Bias | Half-hours |
|---|---|---|---|
| Under 2 hours | 0.72 GW | +0.27 GW | 1,762 |
| 2 to 6 hours | 0.91 GW | +0.55 GW | 2,794 |
| 6 to 12 hours | 1.05 GW | +0.19 GW | 2,392 |
| 12 to 24 hours | 1.87 GW | −1.10 GW | 5,493 |
| 1 to 2 days | 1.90 GW | −1.42 GW | 4,027 |
NESO's transmission demand forecast compared with outturn demand from Elexon. The day-ahead forecast has tended to come in below what was actually used.
Why this matters
Forecasts drive real decisions: when smart chargers start, when batteries fill, and how NESO plans reserves. Showing how good they are, and whether they lean high or low, helps you judge how much to trust the cleanest and cheapest windows ahead.
Common questions
How accurate is the UK carbon intensity forecast?
The table on this page shows the average error for each forecast horizon over the last 30 days, worked out from every forecast the Carbon Intensity API published and the intensity that followed.
How accurate is the electricity demand forecast?
NESO's day-ahead demand forecast is usually within a gigawatt or two of the outcome. We publish its average error and whether it tends to run high or low.
What do mean error and bias mean?
Mean absolute error is the typical size of a miss, ignoring direction. Bias is the average signed miss: positive means the forecast tends to be too high, negative too low.
How do you measure it?
Every 30 minutes we save each forecast as it is published, then compare it with what actually happened once the data settles. Day-ahead means the latest forecast made at least 24 hours before each half-hour.