Offshore wind farm CFD modelling: Uncertainty Quantification and Polynomial Chaos

Abstract

Wind energy will play an essential role in the fight against climate change. By 2050 it is expected to be about a quarter to one third of the total electricity generation. One of the main disadvantages of wind energy is its high variability and low predictability, influenced by physical phenomena at a wide range of time and length scales. The chaotic nature of wind limits the ability of engineering models to predict the performance of wind farms. Furthermore, as wind turbines and wind farms continuously increase in size, thereby increasing their contribution to the power generation industry, the need to better understand the aerodynamic interaction between wind turbines and the atmospheric boundary layer has also increased. Computational fluid dynamics has become an essential tool to enhance our understanding of wind turbine aerodynamics, however, uncertainties are usually overlooked, due to the high computational cost and the lack of characterisation of the different sources of uncertainties. This thesis presents the development of a new computational framework for uncertainty quantification in offshore wind farms. Uncertainty quantification has been identified as one of the key research challenges in the wind energy industry and this work aims to provide a tool that facilitates the propagation of uncertainties in CFD models of wind farms. It is expected that this tool can help to increase our understanding of the wind energy physical system by increasing the amount of information obtained from CFD models providing greater insights and improving the accuracy and confidence on their predictions. The framework implemented integrates the generalized polynomial chaos method (gPC) with OpenFOAM, where a non-axisymmetric actuator disk model (ADM) has been implemented. The ADM was validated against MEXICO and NASA Ames NREL-Phase-VI experiments, and other state-of-the-art numerical models. The framework has been named gpcADM and it has been tested with relatively simple wind turbine arrays considering inflow parameters as random variables. gpcADM captures the response of the system and provides probability density functions for any quantity of interest with a reduced number of deterministic evaluations compared to other traditional sampling strategies.

Date of Award	1 Aug 2021
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Imran Afgan (Supervisor) & Tim Stallard (Supervisor)