Author:

Abstract:

Wind energy is one of the strongest means available to us today to counteract the climate crisis. To facilitate growth in the sector, the underlying technologies have recently witnessed increased research interest in areas such as materials, control and turbine aerodynamics. This has lead to the development of massive mega-watt scale wind turbines, which are often placed together in clusters called wind farms due to economic benefits. However, with increase in turbine and wind farm size, wake overlap effects arising from upstream turbine wakes impinging on downstream turbines lead to a reduction in wind farm efficiency and increase in fatigue of downstream turbine components. To counteract these detrimental turbine--turbine aerodynamic interactions within large farms, coordinated wind farm control strategies such wake steering and axial induction control have emerged as potential solutions to improve overall farm performance. The optimal control set points to realize these strategies are generally determined using simple numerical models due to the large number of simulations required. Unfortunately, their validation through full scale testing of the obtained optimal set points is not always feasible, as experimental campaigns on wind farms can be difficult and expensive to execute. Hence, the current dissertation aims to develop and validate a virtual testing environment which can be used as a reliable and reusable substitute for field testing. By utilizing a large eddy simulations flow solver, coupled with engineering wake models and aeroelastic representation of wind turbines, the current work aims to aid in the development, testing and wide scale adoption of coordinated wind farm control techniques. In the first step, a virtual environment comprising of an in-house large eddy simulation code is utilized to develop a database of atmospheric boundary layers. These precursor inflows are in turn used to develop a database for a reference wind farm employing greedy control. The reference database, comprising of detailed turbine performance metrics in terms of power production, control action and structural loading, serves as a baseline for quantifying the advantages and disadvantages of various coordinated farm control strategies throughout the course of this dissertation. However, before utilizing the virtual environment for testing coordinated wind farm control, the environment itself must first be compared against field measurements to validate the flow and structural solvers. To this end, a data set obtained from a measurement campaign conducted at the Lillgrund offshore wind farm is utilized. Atmospheric conditions measured at the Lillgrund site through lidars are recreated in the flow solver using a novel scaling and shifting approach, and the resulting farm performance of the Lillgrund wind farm simulated in the numerical solver is compared against SCADA and turbine load measurements. Next, the now validated virtual environment is used in open-loop control studies, focusing on two of the most promising coordinated control strategies, i.e. wake steering and dynamic induction control. Through a combination of previously obtained optimal induction settings and an engineering wake model to determine optimal wake steering set points, the virtual wind farm and reference database are used to evaluate the performance of the two strategies. Weaknesses of the open-loop control methodology are also highlighted by comparing the results against wake model predictions. Furthermore, to increase confidence in the obtained results and the numerical tools used in the study, the models are benchmarked against similar tools from industry and academia. Finally, a closed-loop control framework is developed which incorporates online model calibration to improve upon the weaknesses identified in open-loop control. By utilizing measurements, which in our case are obtained from the virtual farm environment as a substitute for a real wind farm, the wake model parameters are tuned to better represent the field conditions and improve its accuracy. A load look up table is also utilized to include structural fatigue in the closed-loop framework, to enable simultaneous power and fatigue optimizations. Benefits of closed-loop control are also exhibited by comparing its performance against open-loop control for dynamic scenarios with turbines shutting down due to downtime or failure.