Retrospective de‐trending of wind site turbulence using machine learning

This paper considers the removal of low-frequency trend contributions from turbulence intensity values at sites for which only 10-min statistics in wind speed are available. It is proposed problem be reformulated as a direct regression task, solvable using machine learning techniques conjunction with training data formed measurements underlying (non-averaged) Once trained, models can de-trend have been retained. A range tested, cases linear and filtered approaches to de-trending, 14 sites. Results indicate this approach allows excellent approximation de-trended distributions unobserved sites, providing significant improvements over existing recommended method. The best results were obtained Neural Network, Random Forest Boosted Tree models.