Data-driven modeling of rotor-induced velocities in sheared inflow

Variations in wind speed and direction affect local thrust and induced velocities over the rotor, impacting power production. To predict power and loading, wind turbine design and control frameworks often leverage blade element momentum (BEM) modeling, where actuator disk momentum modeling for induced velocities is coupled with blade element modeling to predict lift and drag forces on rotating airfoils. The conventional method for estimating induction in BEM modeling is one-dimensional momentum theory, which is derived under the assumption of uniform inflow. Consequently, this theory is limited in its ability to account for the aerodynamic effects of shear that dictate local wind speed over the rotor and that modify turbine power production. We use large eddy simulations (LES) to model turbine rotors, using rotor models that capture turbine rotation and blade airfoil properties, to elucidate the aerodynamic interactions between the rotors and inflow wind conditions. Using LES, we identify in which regimes the relationship between induction and thrust coefficient is insufficiently captured by classical one-dimensional momentum theory. Further, using our LES data, we develop a data-driven model for the relationship between thrust coefficient, wind shear, and the induced velocities. We then evaluate the model against additional LES data not seen during model training. Finally, we compare the results of the data-driven model against several existing models, including classical momentum theory.