Development of reduced order models for wind farm control

Background: Wind energy is one of the leading renewable energies that take part in the current transition from the usage of traditional fossil-fuel energies to green ones. The potential and development of wind power is increasing constantly. Therefore, continuous studies are being carried out in order to improve the power extraction of wind farms. Study: This Master’s Thesis aims to obtain multiple-Input-multiple-Output Reduced-Order Models (IOROMs) that are able to capture the main dynamics present in wind turbine wake flows within wind farms during transients and during operation. This dynamics can be excited via different inputs, such as wind turbine’s yaw angle, generator torque, pitch angle, among others. In this work, the study will be conducted by analyzing the response of the system to yaw angle variations, although the same procedure explained is valid for other inputs. The models developed are also mainly intended for capturing the relation between output magnitudes, such as wind turbine power output, bending moments, lateral forces, etc., and the given inputs to the system. Beside, order reduction is key to this work, since the models are expected to reproduce with acceptable fidelity high computationally costly simulations. The study has been conducted based on CFD (Computational Fluid Dynamics) simulation data that are considered as the starting point of the work. Results & Conclusions: The models developed in this work present considerable agreement with respect to the results obtained via CFD simulations. Owing to the symmetry that the yawing motion presents for the system, the operational range has been divided into two sides, namely the positive yaw angles side and the negative yaw angles side. The results for both cases are successful in terms of flow and power output reconstruction. Therefore, the models can predict within seconds the behavior of wind farms whose performance takes days to be known through CFD simulations. This work also presents several improvements as future upgrade for reduced-order models obtained from CFD data. Some of these improvements deal with operational ranges, sample time, and data collection, inter alia.