Speaker
Description
Accurate day-ahead wind power forecasts are essential for wind farm operation strategies and grid capacity planning. This presentation demonstrates probabilistic day-ahead wind power forecasting using gradient boosting trees. We compare three probabilistic prediction methods - conformalised quantile regression, natural gradient boosting and conditional diffusion models - combined with tree-based machine learning. Validated with four years of data from all wind farms in the Belgian offshore zone, we find that these methods outperform deterministic engineering approaches (power curve or analytical wake model), with an improved point forecast accuracy of around 3.7 % of installed capacity compared to the analytical wake model. Considering the three probabilistic prediction methods, the conditional diffusion model is found to yield the best overall probabilistic and point estimate of wind power generation.
