Aerodynamic Simulations For Wind Turbines In Flow Separation And Stall

It is a great challenge to predict aerodynamic performance of wind turbines in flow separation andstall for various calculation methods including computational fluid dynamics (CFD). This challengewas highlighted in National Renewable Energy Laboratory (NREL) Unsteady AerodynamicsExperiment Phase Ⅵ. It is of significance for accurate aerodynamic simulations for wind turbines tobreakthrough difficulties encountered in simulations on the Phase Ⅵ blade in flow separation andstall. In the present study, numerical simulations are performed for the S809airfoils and the Phase Ⅵblade by solving RANS equations using unstructured finite volume scheme. Great efforts are made tosimulate the aerodynamic performance in flow separation and stall more accurately in the followingaspects:The error correlation between simulations on the Phase Ⅵ blade and on the S809airfoil isrevealed. Ignoring transition and inaccurate turbulence simulation are recognized as the main twosources responsible for the error in conventional full turbulence simulations. Two measures are takento reduce the simulation error. One concerns about the transition prediction. The occurrence of deepstall is predicted more exactly as the γ-Reθtransition model is introduced, capturing the laminarseparation bubble near the leading edge. The other measure is to calibrate a closure constant denotedby β*of SST turbulence model. It is found there exists an optimal value of β*appropriate for theentire angle of attack range and a certain Reynolds number range, while only a small amount ofexperimental data at individual angle of attack is needed as a reference for obtaining the optimal value.Simulations on both the airfoil and the blade are significantly improved throughout the light stallstage as the optimal value of β*is adopted.A wind tunnel test was carried out for a1/8scaled model of the Phase Ⅵ blade. Globalskin-friction lines on the model were measured using the luminescent oil-film method, and intuitiveflow pictures were obtained for validating the present numerical method. In addition, a method forcredibility judgment of the measurement results was proposed. The scaled model showed a significantdifference in aerodynamic performance from the original full size blade. Indeed, laminar flowseparation occurred near the trailing edge as the angle of attack was only0o due to such a lowReynolds number for the scaled model. Full turbulence simulations fail to predict the laminar flowseparation while transitional simulations do better. It highlights the advantage of the transitionalsimulation method in conditions of low Reynolds numbers.Looking into the improved simulation results, the mechanism of the three-dimensional rotational effect is revealed. The separated vortices on the rotational blade show strong three-dimensionalcharacteristics under the action of centrifugal force. It is found that stall delay is not totally equivalentto separation delay but is mainly caused by shape flattening and size reduction of the separationvortices. Based on above findings, a new stall delay model is established for a better correction on theincrease of aerodynamic force due to the three-dimensional rotational effect. The application of themodel into actuator line method improves the numerical results significantly in stall conditions.