An ML-based wind turbine blade design method considering multi-objective aerodynamic similarity and its experimental validation
Siyao Yang, Kun Lin, Annan Zhou
Abstract
Model test is an essential technique to study the aerodynamic performance of wind turbines. To overcome the poor aerodynamic performance of scaled models caused by the scaling effect, this study proposes an innovative blade design method for scaled model testing based on machine learning (ML). The method achieves satisfactory similarity between the thrust and power coefficients under multiple operating conditions of the model and prototype. Furthermore, a case study of the NREL 5-MW wind turbine is carried out with wind tunnel tests to validate the effectiveness of the proposed method. Obtained results suggest that the aerodynamic performance of redesigned blade closely mirrors that of the prototype under multiple operating conditions, reaching 97.59 % (thrust) and 97.87 % (power) coefficients of the prototype at the rated operating condition, respectively. With this technique, aerodynamic performance similarities between the redesigned blade and the prototype can be enhanced, contributing to more accurate scale model testing.