Aerodynamic shape optimization of horizontal axis wind turbines was performed to minimize the cost of energy. The optimum blade was obtained by modifying the blade section contours at selected blade radial stations. For the optimization, the design variables were defined as the coordinates of the blade section contours perpendicular to the chord. An aerodynamic performance database was constructed for those blade configurations defined by random combinations of design variables selected within a given range by Latin-Hypercube sampling. An artificial neural network was applied to derive the approximate functions between the design variables and the aerodynamic performance of the database. By utilizing the approximate functions, a genetic algorithm was adopted to search for an optimized blade. To construct the database, a CFD flow solver based on unstructured meshes was utilized. To consider the effects of aeroelastic deformation of the structurally flexible blades, a coupled CFD-CSD method was also adopted. The applications were made for the NREL phase VI and NREL 5 MW reference wind turbine rotor blades. After optimization, the cost of energy was reduced by 0.82% for the NREL phase VI rotor blade and one percent for the NREL 5 MW reference wind turbine blade, respectively.
Read full abstract