High fidelity CFD simulation of horizontal axis wind turbines

The aim of the thesis is to carry out high fidelity computational fluid dynamics, CFD, simulations on the DTU 10MW wind turbine in order to understand (1) what is the best way for generating grids, (2) what is the optimum turbulence model to mimic the aerodynamics behind the turbine and (3) what are the effects of the cell size extruded from the blade, thus on the use of wall functions, on the performance computation. The first part of the thesis deals with several studies done on the NACA 0012 airfoil profile. The latter were done with the aim of familiarizing with the solver Open- Foam and with the grid generator Pointwise and of shallowly forecast, from the airfoil, what the behaviour of the turbine simulations would be by using similar settings. From the latter we chose to use SIMPLE-Consistent algorithm (since it is faster and more robust than SIMPLE), to study only two turbulence models on the turbine (Spalart-Allmaras and k-omega-SST are the most accurate), to use upwind divergence scheme for turbulence variables and second order upwind for the velocity and to employ wall functions since results were not so different from the ones evaluated with a fine mesh and a direct solution. The wind turbine grids were generated with hybrid technique employing struc- tured mesh in the region closer to the blade and unstructured mesh in every other region of the domain. Most of the generation operations were done manually since the use of automatic tools caused high non-orthogonality and high skewness problems. In the end, four grids with different first wall layer lengths were created in order to study the effects of wall functions. Good results were achieved since our grids have less than the half of the elements of fully structured grids found in the literature and the performance obtained from the CFD analysis were really similar to the literature ones. In order to find out which is the most accurate turbulence model, both steady state and transient simulation were performed on the four grids, employing the iii abovementioned two models. The power and thrust obtained at different wind speeds were compared with the ones gathered from other three universities. Since the turbine is still in the design phase we do not have any experimental result. The two models carried out similar results. It was not possible to unambiguously state which turbulence model behaves better since results were not always better with one model. However, on average, k-omega-SST model showed better prediction of power and thrust and more accurate pressure and velocity fields, although it is slower in convergence with respect to Spalart-Allmaras model. The conclusion inferred was that it is better to use Spalart-Allmaras model at the very beginning of the design phase and to use k-omega-SST when the geometry of the blade is almost fully defined. The use of wall functions is definitely suggested for this kind of simulations. Spalart-Allmaras model is almost insensitive to wall functions, thus coarse grids can be correctly employed without any loss in the performance prediction. K- omega-SST are a bit more sensitive when we deal with higher wind speeds, but the differences are in the order of 1%. The thesis was carried out at the Statik Department, in the Technische Universitaet Muenchen under the supervision of Prof. Bletzinger thanks to the ITALDESIGN- GIUGIARO project.