Beschreibung
Current estimates assume that the share of wind energy in the global energy mix will increase fivefold until 2050. Although wind energy has left its early stages of development and already is a well-established source of electricity, additional research is needed to face future engineering challenges that arise from the deployment expansion of this technology. In the field of wind turbine rotor aerodynamics, the current industrial load calculation almost entirely relies on low-fidelity models, e.g. methods that are based on the blade-element-momentum theory (BEMT). Due to their low computational cost they provide a valuable tool for the iterative design process. However, their prediction performance strongly depends on the fidelity of semiempiric submodels. Since computing-intensive high-order methods are not applicable in industry to tackle the uncertainty caused by the use of these models, a new generation of tailor-made engineering tools which represent a compromise between computing time and accuracy is needed. Vortex panel methods are particular promising to close this gap. To gain further insights on their potential to enhance the aerodynamic load calculation of wind turbine rotors, a three-dimensional vortex panel method was implemented, which can be augmented with a vortex particle method. Based on experimental and numerical findings for the flow over a three-bladed rotor, the aerodynamic prediction performance of the implemented vortex panel method is assessed for axial and yawed inflow conditions of the rotor. For the axial inflow of the rotor, it is demonstrated that the predictions of the panel method are on a par with the findings of BEMT-based approaches and Reynolds-averaged Navier-Stokes (RANS) solvers, if the missing influence of the boundary layer is taken into account by an estimation. At the same time, the required computing time using the vortex panel method is in the range of O(100) to O(101) hours on a standard desktop computer, which is a fraction of the time of higher-order methods using the same hardware setup. Furthermore, it is demonstrated for yawed inflow conditions of the rotor that the implemented method inherently captures the three-dimensional character of the flow around the rotor without the need for a semi-empiric inflow model. To accelerate the implemented vortex method, the graphic processing unit (GPU) of a modern consumer graphics card is used. In addition, a novel pseudo-particle method is presented which overcomes the n-body problem characteristics of vortex methods. This combination allows to conduct complex simulations on a standard desktop computer without the need for a multi-core cluster. The findings of this thesis show that vortex panel methods represent perfectly tailored tools to close the gap between current low and high-fidelity methods applied in the field of computational aerodynamics. Therefore, they offer a huge potential to enhance the industrial aerodynamic load calculation of wind turbine rotors.
