RANS Computation of 360° Polars for Wind Turbine Airfoils

Kurzfassung

Horizontal axis wind turbines have to withstand a wide range of atmospheric operating conditions from turbulent inflow to stand still. This can lead to airfoil incidence angles from any direction including reverse flow. Blade element methods require aerodynamic airfoil coefficients for the computation of the aerodynamic loads on the rotor blade. The state of the art method to acquire these aerodynamic coefficients is to take the polars from -30° to +30° from either wind tunnel experiments or CFD calculations and to extend it by the analogy of the flat plate (e.g. the post stall model from Montgomery1). One problem of this approach is the discontinuity at the junction between the measured and the extended data due to different levels of physical modelling. Furthermore the flat plate analogy does not take into account effects like surface curvature or camber. For the data taken from wind tunnel experiments large test models and test sections are needed in order to meet the demanding requirements of the high Reynolds numbers typically encountered on modern wind turbine rotor blades. In contrast to that, CFD methods are less time consuming, especially when a lot of different airfoil geometries have to be investigated. At high angles of attack (AoAs) the flow separation from the airfoil leads to unsteady aerodynamical effects. Higher-order CFD methods such as Large Eddy Simulation and Direct Numerical Simulation would disclose a precise aerodynamic but exceed an appropriate effort for the whole range of AoAs and airfoil types. Due to the time efficiency aspect the steady RANS computation was chosen as simulation method for the DLR TAU code. The goal was to find one robust CFD setup with one mesh topology that allows the computation for the whole range of AoAs and of all airfoils in order to get smooth slopes in the aerodynamic coefficients. For the present paper five airfoils of the FFA-W3-xxx series2 used on the DTU 10MW wind turbine3 were chosen. The airfoils have a relative thickness from t/c=24.1% up to t/c=60% whereby some airfoils are equipped with a gurney flap. The flow conditions vary from Re=6x106 to Re=12x106 and M=0.058 to M=0.265 and, therefore, are in accordance with the rated operating condition of the rotor. Two different turbulence models (one-equation SA-neg and two-equation Menter-SST) were tested, whereas the SA-neg model was favored for the final simulations. The SA-neg model was originally developed for under-resolved grids and unphysical transient states resulting in a more stable solution in difficult flow conditions. It has been found out that the laminar-turbulent transition may not be neglected as it has an important impact on the suction peak even at high AoAs. The transition prediction has been performed by the eN method using the linear stability theory. Particularly for thick airfoils with large nose radius the consideration of transition affects the lift coefficient up to 0.4 and the drag coefficient up to 0.2. In few cases the flow solution did not converge. The respective points were excluded and interpolated by a higher order polynomial. This methodology was compared to the classical approach by which the aerodynamic coefficients in the range from -30° to +30° were computed by CFD and extended by the analogy of the flat plate. It can be shown that the presented methodology resolves physical effects in the coefficient shape in a more sophisticated manner. Furthermore the data are more consistent than the classical approach.