This work shows the development of a tool for wind farm layout optimization based on computational fluid dynamics (CFD) simulations of the atmospheric wind flow and inter-turbine interference. Due to the high computational power required to simulate a whole wind farm using the complete geometry of the wind turbines, simplified models were developed to represent the turbine behavior, and the most commonly used model is the actuator disk model and its variations. The procedure for wind turbine behavior evaluation using a CFD model was implemented in the OpenFOAM software, and this model was coupled with the Dakota optimization toolkit. A genetic algorithm was selected for the optimization task due to its robustness and the characteristics of the problem solved. With this new tool in hand, three different terrain cases, with growing complexity, were tested considering different numbers of turbines on a cylindrical domain in order to achieve the best wind farm layout in terms of annual energy production that respects the imposed physical restrictions. As expected, the layout efficiency decreased as turbines were added to the domain, meaning that wake losses are introduced in this process. For the simpler domains, we observed that this efficiency decrease was approximately linear with respect to the number of turbines. Complex phenomena were captured during the optimization, such as wake deflection, different wake recoveries depending on the wind speed and interaction between the disturbances in the flow field caused by the terrain and by the turbines. These phenomena are not observed when using the tools available in the wind market and are seen as an improvement in the field of wind farm layout optimization, yielding more accurate results, and should decrease the level of uncertainty in the design of a wind farm.

这项工作显示了基于大气风流和涡轮机间干扰的计算流体动力学（CFD）模拟的风电场布局优化工具的开发。由于要使用风力涡轮机的完整几何结构来模拟整个风力发电场需要很高的计算能力，因此开发了简化模型来表示涡轮机的性能，最常用的模型是执行器盘模型及其变体。在OpenFOAM软件中实施了使用CFD模型进行风轮机性能评估的程序，并将该模型与Dakota优化工具包结合使用。由于遗传算法的鲁棒性和所解决问题的特征，因此选择遗传算法进行优化。有了这个新工具，复杂性不断提高的三种不同地形情况，在考虑圆柱形区域上不同数量的涡轮机的情况下进行了测试，以实现在遵守强制性物理限制的年度能源生产方面的最佳风电场布局。正如预期的那样，随着将涡轮机添加到域中，布局效率降低了，这意味着在此过程中引入了尾流损失。对于较简单的域，我们观察到效率的降低相对于涡轮机的数量是近似线性的。在优化过程中捕获了复杂的现象，例如尾流偏转，取决于风速的不同尾流恢复以及地形和涡轮机在流场中的干扰之间的相互作用。