Modeling and uncertainty quantification for cooperative wind farms at utility-scale

Wind turbines located in wind farms are operated to maximize only their own power production.
Individual operation results in wake losses that can reduce energy 10-20%. Wind farm flow control has
demonstrated potential to increase collective power production. To achieve the maximum farm power
production, the models used for wind farm decision-making must be both accurate and computationally
efficient. The potential for flow control depends, in part, on the power reduction of yaw misaligned
turbines. In the atmospheric boundary layer (ABL), the sheared wind speed and direction may change
significantly over the rotor area, resulting in a relative inflow wind to the blade airfoil which depends on
the radial and azimuthal positions. In order to predict the power production for an arbitrary yaw
misaligned turbine based on the incident ABL velocity profiles, we develop a blade element model which
accounts for wind speed and direction changes over the rotor area, and the model is validated using
experimental data from a utility-scale wind farm. The blade element model is coupled with an
aerodynamic wake model to establish a collective flow control model. Leveraging the flow control
model, we designed a physics-based, data-assisted wake steering control method to increase the power
production of wind farms, which utilizes data assimilation and gradient-based optimization. The method
was first validated, demonstrating that it predicts the true power maximizing operation, and then tested
in a multi-turbine array at a utility-scale wind farm, where it statistically significantly increased the
energy production over standard, individual operation. Collective control can increase the generation
potential of wind farms through software modifications, without additional turbines or hardware. To
enable reliable decision-making for cooperative wind farm design and control under inflow wind
uncertainty, we develop a framework for efficient uncertainty quantification (UQ) of computationally
intensive numerical ABL flow models. We validate the methodology for efficient parameter UQ in an
idealized general circulation model (GCM) and in large eddy simulations (LES) of the stratified ABL.