Blade Element Momentum Theory Coupled with Machine Learning to Predict Wind Turbine Aerodynamic Performances

So far, the simplest aerodynamic method for wind turbine design and optimization is the Blade Element Momentum (BEM) theory. BEM method needs the lift and drag coefficients as inputs in its algorithm to predict the aerodynamic performance of wind turbine rotors. These coefficients are commonly obtained from numerical simulations or experimentally. In both cases, important time and resources are necessary. In this work, an alternative technique based on the machine learning, namely Artificial Neural Network (ANN) is used and validated. The main objective of this study is to develop and optimize an ANN architecture for predicting aerodynamic performance in general wind turbine rotors. Firstly, a parametric study including more than 2.10 E13 training data points was carried out to predict the aerodynamic coefficients of airfoils. This parametric study takes into account many ANN features such as the number of needed hidden layers, the number of neurons in each layer, the impact of activation functions, the impact of input models, and the learning techniques. After that, the optimized ANN model has been coupled with the classical BEM algorithm to predict the aerodynamic performance of wind turbine rotors, specially in cases where airfoil data is not available. Finally, it has been shown that this coupled BEM-ANN approach provides accurate results in a very short time. Consequently, this technique can be considered as a powerful tool and a fast model to predict the wind turbines performance.