Forecasting
How to Build Solar and Wind Generation Forecasts
Because solar and wind output depend on weather, forecasting models must combine meteorology with plant behaviour. A useful forecast provides not only MWh but also an uncertainty range and update time.
Align weather and site data
Irradiance, cloud cover, temperature, wind speed and direction must align with telemetry by coordinates, height and timestamp. Sensor faults, icing, shading and maintenance events should be labelled separately.
Power curves and capacity limits
Model output should respect DC/AC ratio, inverter or turbine power curves, installed capacity and available capacity. Training on curtailed periods as if they were normal generation can create persistent underforecasting.
Manage error and uncertainty
Track MAE and normalised error separately by horizon, and consider weighting hourly errors by cost or imbalance impact. Prediction intervals help portfolio and trading teams identify riskier hours before delivery.
Sources
Related articles
More STR Energy guides to continue exploring the topic.