Deep learning predictions of unsteady aerodynamic loads on an airfoil model pitched over the entire operating range

For the design and certification of wind turbines, it is essential to provide fast accurate unsteady aerodynamic load prediction models for whole operational range angle attack, up 180° vertical-axis 90° horizontal-axis turbines. This work describes a computationally efficient forces model based on deep learning approach, namely bidirectional long short-term memory (BiLSTM) algorithm, an airfoil pitched over full angles attack 180°. No has been developed capture at high attack. Novel features operating conditions steady polars are used as inputs BiLSTM model. Direct measurements NACA 0021 were conducted reduced frequencies 0.075 Reynolds number 120 000 in open-jet tunnel testing. The vary significantly from values pitching amplitudes post-stall angles, which, if not accounted when simulating turbine performance, would result inaccurate predictions. Furthermore, revealed effect vorticity development shedding under forward reverse flow conditions. capable capturing underlying physics extreme