MW
- P10
- P50
- P90
Live Predictions
Loading live data…
How does this compare to NESO's own forecast?
This project's headline result: blending our model with NESO's day-ahead forecast (NDF) beats NDF alone by ~3.55% on out-of-sample testing (weight ≈0.17 on our model). That combination only exists at NDF's own published cardinal-point times, so it's shown here separately rather than folded into the 1-day tile above.
- NDF cardinal point
- –
- NDF forecast
- –
- Our forecast
- –
- Combined (weight 0.17)
- –
How has this performed over the last year?
Daily-averaged actual demand vs. NESO's day-ahead forecast (NDF) vs. our blended forecast, for the trailing 12 months. The "our forecast"/"blended" lines use actual historical weather (not a real forecast) as an input, and the two rightmost points a few days apart reflect actual demand's own reporting lag — see the note below the chart for both caveats in full.
Actual demand
NDF (NESO)
Our blend
Loading…
Forecast error vs. actual (NDF and our blend)
Drag on the chart to zoom into a date range; hover for exact values.