Blade element momentum theory is extensively used to design and characterize the performance of the wind turbine. Aerodynamic characteristics of the airfoils used in the blades of the wind turbine are one of the crucial parameters of the blade element momentum theory. The aerodynamic characteristics of airfoil can be significantly influenced by Reynolds number besides the angle of attack. Thus, an inadequate consideration of Reynolds number for aerodynamic characteristics of the airfoil can result in discrepancy in the optimum design and performance evaluation capability of the blade element momentum theory. Several mathematical models, and semi-empirical relations to parameterize the aerodynamic characteristics of two-dimensional airfoil in blade element momentum theory, inherently lack dependency on Reynolds number and thus open a scope for uncertainty. An artificial neural network-based model is proposed to predict the lift and drag coefficient of airfoil as a function of not only the angle of attack but also the Reynolds number. A series of six in-house developed airfoils for small wind turbine have been considered for the present study. The computational fluid dynamic results of the airfoil with a range of Reynolds number (100,000–2,000,000) and angle of attack (0°–20°) were utilized to develop the model. A high coefficient of determination and low root-mean-square error of the developed models for a test dataset suggests the robust capabilities and effective topology of the artificial neural network-based model to predict the lift and drag coefficient of the airfoils with respect to a given angle of attack and Reynolds number. The developed model can then be used to replace the traditional analytical or semi-empirical model for mathematical representation of airfoil in the blade element momentum theory and thus reduce the uncertainty on account of inadequate consideration of Reynolds number for aerodynamic characteristics of airfoil in design and performance evaluation.

Graphic abstract

中文翻译：

---

基于人工神经网络的元模型预测小型水平轴风力发电机的二维翼型气动特性

叶片元素动量理论被广泛用于设计和表征风力涡轮机的性能。风力涡轮机叶片中使用的翼型的空气动力学特性是叶片元件动量理论的关键参数之一。除了迎角，雷诺数还会显着影响翼型的空气动力学特性。因此，对翼型的空气动力学特性没有充分考虑雷诺数会导致叶片元件动量理论的最佳设计和性能评估能力出现差异。几个数学模型和半经验关系，用于在叶片单元动量理论中对二维翼型的空气动力学特性进行参数化，固有地缺乏对雷诺数的依赖性，因此为不确定性开辟了空间。提出了一种基于人工神经网络的模型，以预测机翼的升力和阻力系数，不仅取决于迎角，而且还取决于雷诺数。本研究考虑了一系列由六个内部开发的小型风力涡轮机用机翼。利用翼型范围为（100,000–2,000,000）和迎角（0°–20°）的翼型的计算流体动力学结果来开发模型。测试数据集的已开发模型的高确定系数和低均方根误差表明，基于人工神经网络的模型的稳健能力和有效拓扑可预测机翼相对于翼型的升力和阻力系数给定迎角和雷诺数。然后，可以将开发的模型用于代替传统的解析模型或半经验模型，以在叶片元素动量理论中对机翼进行数学表示，从而由于在设计和性能方面未充分考虑雷诺数来考虑机翼的气动特性，从而减少了不确定性评估。

图形摘要