Live Forecast
Actual German renewable generation (wind + solar) vs ENTSO-E day-ahead vs our model.
Last 48h actuals + next 24h model forecast. Model MAE/MAPE use 24h-ahead replay vs actuals.
Model Card
Transparency surface for the current LightGBM model, training metrics, and top features.
Live Skill Monitor
Why the green model line deviates from actuals and ENTSO-E — bias, MAE, and skill over time.
Scenario Forecast
What-if analysis: adjust wind and solar capacity multipliers to see how total generation changes vs the baseline forecast.
Backtest Metrics
Evaluates forecast accuracy over the selected date range using standard error metrics (MAE, RMSE, MAPE) and compares against a naive baseline.
AI Insights
Natural-language explanations of forecast drivers and model behavior.
How it works
End-to-end view of the forecast pipeline — from public data to model, skill, and explainability.
ENTSO-E & Open-Meteo
Ingests German wind onshore, wind offshore, solar PV, and total load from the ENTSO-E Transparency Platform, enriched with weather features from Open-Meteo.
LightGBM quantile regressors
Trains point and quantile regressors to output both a central forecast and 80% prediction intervals for every 15-minute step.
Live skill vs ENTSO-E
Benchmarks the model against the official day-ahead forecast with MAE, RMSE, MAPE, coverage, and skill score over live and historical windows.
Ask the Forecast
A tool-calling LLM agent answers questions from live forecast, actuals, weather, model metadata, and a small knowledge base — useful for exploring drivers in plain language.
Export Data
Download raw generation data at 15-minute resolution from ENTSO-E for the selected date range.