Vortex-model-based Multi-objective Optimization of Winglets for Wind Turbines using Machine Learning

Abstract Different Design Driving Load constraints (DDLs), are explored in this work to determine under which and conditions a winglet can have an added value the wind turbine blade design. Multi-objective Bayesian optimization is used maximize rotor’s power production while minimizing flapwise DDLs. Surrogate models, created using machine learning techniques such as Gaussian Processes Neural Networks, combination with acquisition function, what designs should be evaluated by lifting line model AWSM, goal obtain that lie on Pareto front of two or more objectives. The recent Networks surrogate were able find Pareto-front most effectively work. Furthermore, results show different DDL led designs, noticeable differences between upwind downwind designs. Winglet found increase without increasing thrust, root bending moment at radial locations blade. A was when introducing sweep