For wind turbine design optimisation, it is necessary to have a fast model to evaluate the loads at each design iteration. However, current IEC fatigue and ultimate load cases require time-dependent aeroelastic simulations that are too computationally expensive. Therefore, Barlas, Göçmen, and Riva (2024) investigate how inexpensive machine-learning models of the blade and component loads can be trained based on aeroelastic simulation.

The authors use the IEA 3.4 MW wind turbine as a reference and the blade length, tip-speed ratio, and pitch angle as design variables. One hundred wind turbine design alterations are simulated using the Aeroelastic code HAWC2, and the resulting data is used to train the load models. The results show that the surrogate load models are accurate. Barlas, Göçmen, and Riva (2024) suggest expanding it with other design variables and site-dependent wind conditions so that the model can serve for both wind turbine and wind farm layout optimisation.

Barlas, T, T Göçmen, and R Riva. 2024. “Development of a Machine Learning Model for Wind Turbine Fatigue and Ultimate Loads Based on Static Loads.” Journal of Physics: Conference Series 2767 (5): 052009. https://iopscience.iop.org/article/10.1088/1742-6596/2767/5/052009