A from-scratch machine learning model that forecasts Great Britain's national electricity demand 30 minutes, 1 hour, 1 day, and 7 days ahead — with calibrated P10/P50/P90 uncertainty bands, not just a single number.
Every prediction on this site is computed on demand, from data fetched right now: NESO's rolling demand feed, a live Open-Meteo weather forecast, and NESO's current day-ahead forecast — the same pipeline shape as the model's own training data, not a canned demo.
Alongside a single best-guess number, each horizon reports a P10–P90 range: a forecaster's honest statement of "here's a band I'm 80% confident contains the real value," not false precision.
The longer the horizon, the more the forecast leans on a live weather forecast rather than certainty — and the 1-day tile is shown alongside NESO's own published forecast, not in place of it. See the report for the full limitations.