Uses day-ahead wholesale prices from the Nord Pool N2EX market to predict Agile rates. A rolling linear regression is fitted over the most recent 14 days of data to derive the coefficients D (wholesale weight), P (peak surcharge), and a base offset. The formula is price = D × wholesale + P × peak + base, with a 5% VAT applied. Currently fitted at D=2.16, P=12.2, base=-3.8, MAE=1.1p. Available once Nord Pool publishes day-ahead prices (usually by 11am for the next day).

An XGBoost ensemble model (V4) trained on historical Agile rates, wholesale prices, and weather data. It uses separate sub-models for off-peak, peak, and shoulder periods, each with time-of-day features, lagged prices, and weather forecasts (temperature, wind speed from Open-Meteo). Predictions are chained — each slot's forecast becomes an input for the next. Used to fill gaps beyond Nord Pool data availability, typically covering 24–48 hours ahead.