• Wake
• turbulence
• gPC
• Actuator Line Model
• wind turbine

Today the investment and the attention to renewable energies has increased. In order to produce the required amount of energy, wind turbines are clustered in wind farms, in which the distance between the turbines is not enough to allow the recovery of the wake. The wake plays a key role in the power production, because it is characterized by a turbulent flow that contains less kinetic energy, compared to the undisturbed incoming velocity. Thus, a reliable numerical method giving a good prediction of the wake is crucial for wind turbine applications.

A typical model used for the numerical simulation of this type of flows is the so-called Actuator Line Model (ALM). In this method, the real body is replaced by a distribution of forces spread in a cylinder having radius 'e' (spreading parameter), which integral is the total aerodynamic force acting on the body. In previous studies performed on the airfoil NACA0009 at high angles of attack, it has been found that the ALM approach can very well predict the turbulent wake of the real body, which has been simulated with the Immersed Boundary Method (IBM). In particular, for a calibrated value of the spreading parameter, the width and the global amount of turbulence in the wake is in very good agreement with the IBM results. However, it has been also noticed that the ALM wake is significantly below that of the real body.

The aim of this thesis is to further investigate the accuracy and the reliability of the actuator line model predictions for turbulent separated wakes by introducing the aerodynamic moment effects. The enhanced ALM model contains a second spreading parameter for the force couple reproducing the aerodynamic moment. Since a fully deterministic sensitivity analysis of 2 parameters is excessively expensive for LES simulations, we used a stochastic approach adopting the generalized Polynomial Chaos (gPC) expansion.

Finally, the optimal couple of spreading parameters found with the gPC analysis is compared to the optimal set-up of the 'classic' ALM approach without the implementation of the aerodynamic moment.