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

Abstract

In order to reduce the levelised cost of energy, the rotors of wind turbines are increasing in size. To increase the energy yield, wind turbine rotors need to have an innovative tip design; such as winglets. Winglets are used widely in aircraft design; however, they remain mostly absent in state-of-the-art wind turbine design. The low-fidelity wind turbine design models used by industry, such as BEM, are insufficient to capture the full 3D flow physics in such an innovative design. Therefore, high-fidelity methods, such as vortex methods, are becoming more and more important in such a design phase in wind energy research. Single-objective optimisation has been applied in earlier works to maximise power production. Winglets have shown the potential of increasing power production while simultaneously increasing the design-driving loads (DDLs), for example, the thrust or flapwise bending moment. This work focuses on optimisation using machine learning of a winglet that increases power production without increasing DDLs.

A parameterised design for a winglet on a wind turbine is created. Different design constraints, such as DDLs or a diameter constraint, are explored to determine under which constraints and conditions a winglet can have an added value to the wind turbine blade design. Multi-objective Bayesian optimisation is used to maximise the rotor's power production and minimise DDLs. Surrogate models, created using machine learning techniques such as Gaussian Processes and Bayesian Neural Networks, are used in combination with an acquisition function, to determine what designs should be evaluated by the lifting line model AWSM. This has the goal to obtain designs that lie on the Pareto front of two or more objectives. The recent Bayesian Neural Networks were able to find the Pareto front most effectively in this work. Furthermore, the results show that different DDL constraints led to different winglet designs, with noticeable differences between upwind and downwind winglet designs obtained by the optimiser. Both a downwind and upwind winglet were found to be able to increase power without increasing the thrust, root flapwise bending moment and flapwise bending moment at 80% of the rotor radius.