Wind energy plays an increasingly important role in the transition towards a sustainable and low emission energy system. As wind turbines and wind farms continue to increase in size, which becomes comparable to atmospheric boundary layer height, the accurate prediction of wake effects becomes essential for both power production and structural loading. One phenomenon that strongly influences wake behavior is wind veer, defined as the variation of wind direction with height. Under veered inflow conditions, turbine wakes experience lateral displacement and profile skewness, affecting the performance of downstream turbines.

This thesis investigates the ability of engineering wake models to predict the lateral displacement of wakes under veered inflow conditions using Large Eddy Simulation (LES) data for comparison. The lateral wake displacement has been defined as the lateral position of the maximum velocity deficit over the vertical axis. Existing analytical models developed by Abkar, Qian, Narasimhan and Snaiki are evaluated and compared against the LES, while several improvements are examined, including alternative advection formulations, the use of downstream velocity fields, analytical integration techniques and ellipse fitting methods.

Finally, two different physics-guided neural networks are developed. The first estimates a correction factor for the Abkar model. While the second, predicts the parameters of a rotated Gaussian ellipse. The proposed methods significantly improve the prediction of lateral wake displacement and lead to substantially better agreement with both the training and the test (previously unseen) LES cases. These findings contribute to the development of more accurate engineering wake models for atmospheric flows with strong wind veer.
... Read more