Uncertainty quantification of wind turbine airfoil aerodynamics with geometric uncertainty

Abstract An artificial neural network based reduced order model (ROM) is developed to predict the load coefficients and performance of wind turbine airfoils. The trained using a representative database 972 airfoil shapes generated by perturbing design parameters in each 12 baseline airfoils defining commercially relevant modern turbines. predictions from our ROM show excellent agreement with CFD data, 99 th percentile maximum errors 0.03 lift-coefficient, 2 lift-to-drag ratio 0.002 pitching moment coefficient. A Monte-Carlo uncertainty quantification (UQ) global sensitivity analysis (GSA) framework this computationally economical ROM. Using UQ, we observed stall behavior be very sensitive geometric uncertainty, more than 10% deviation lift coefficient associated 5% features. Sobol’s used identify most influencing feature for concentrated at thickness location on suction surface.