Statistical learning for wind power : a modeling and stability study towards forecasting
This work addresses wind power forecasting for energy companies, but it is incremental as it builds on existing machine learning techniques with specific refinements.
The study tackled wind power modeling by comparing parametric models with machine learning algorithms on real wind energy data, finding that CART-Bagging outperformed parametric approaches in stability and performance. It also quantified the impact of deteriorated wind measurements on forecasting and refined predictor selection methods in random forest packages.
We focus on wind power modeling using machine learning techniques. We show on real data provided by the wind energy company Ma{ï}a Eolis, that parametric models, even following closely the physical equation relating wind production to wind speed are outperformed by intelligent learning algorithms. In particular, the CART-Bagging algorithm gives very stable and promising results. Besides, as a step towards forecast, we quantify the impact of using deteriorated wind measures on the performances. We show also on this application that the default methodology to select a subset of predictors provided in the standard random forest package can be refined, especially when there exists among the predictors one variable which has a major impact.