The graphic illustrates the wind energy forecasting process using machine learning. Numerical weather forecasts and live data as well as additional customer data are fed into two machine learning models: one for power prediction and one for event identification. These models can be applied to individual turbines and entire wind farms. The identified events (e.g. shutdowns, icing) are stored in an event database and taken into account in the final forecast. The process leads to a precise forecast of energy production.