Search

The majority of goods transportation vehicles' power is consumed in overcoming aerodynamic drag which arises due to vortex shedding over the bluff rear end of the vehicle. As such, a reduction in pressure drag via feedback control could have significant economic and environmental effects on CO2 emissions, as well as reducing fatigue acting on the body. The difficulty in designing such controllers lies in obtaining models suited to modern control design methods, which are necessarily of much lesser complexity than typical Computational Fluid Dynamics models, or models derived from immediate spatial discretisation of the equations governing fluid flows. It is with obtaining such low-order models that the work presented in this thesis is concerned. A computationally efficient modelling approach which is suited to obtaining low- order models of complex geometry fluid flows is described, whereby the system's overall input-output frequency response is built up by connecting together the frequency responses of a large number of computational node subsystems in an efficient manner, exploiting the inherent structure of spatially discretised PDAEs. In order to choose a suitable formulation of the governing equations – the Navier- Stokes equations – a rigorous analysis of several of the formulations suggested in the literature is presented, whereby the dynamics of different formulations are compared both at the nodal level, and at the full system level for a 2D channel flow (for which a well studied benchmark model exists). In the penultimate chapter, the work of previous chapters is consolidated by apply- ing the modelling technique to a 2D backward facing step flow. Slot jet actuation on the rear edge is assumed, and two separate output configurations are considered. The resulting models are compared to models obtained in a computational system iden- tification study, which prompts an interesting investigation into the dynamics of the common PISO Computational Fluid Dynamics algorithm.

Large wind turbines are subjected to the harmful loads that arise from the spatially uneven and temporally unsteady oncoming wind. Such loads are the known sources of fatigue damage that reduce the turbine operational lifetime, ultimately increasing the cost of wind energy to the end users. In recent years, a substantial amount of studies has focused on blade pitch control and the use of real-time wind measurements, with the aim of attenuating the structural loads on the turbine blades and rotor. However, many of the research challenges still remain unsolved. For example, there exist many classes of blade individual pitch control (IPC) techniques but the link between these different but competing IPC strategies was not well investigated. In addition, another example is that many studies employed model predictive control (MPC) for its capability to handle the constraints of the blade pitch actuators and the measurement of the approaching wind, but often, wind turbine control design specifications are provided in frequency-domain that is not well taken into account by the standard MPC. To address the missing links in various classes of the IPCs, this thesis aims to investigate and understand the similarities and differences between each of their performance. The results suggest that the choice of IPC designs rests largely with preferences and implementation simplicity. Based on these insights, a particular class of the IPCs lends itself readily for extracting tower motion from measurements of the blade loads. Thus, this thesis further proposes a tower load reduction control strategy based solely upon the blade load sensors. To tackle the problem of MPC on wind turbines, this thesis presents an MPC layer design upon a pre-determined robust output-feedback controller. The MPC layer handles purely the feed-forward and constraint knowledge, whilst retaining the nominal robustness and frequency-domain properties of the pre-determined closed-loop. Thus, from an industrial perspective, the separate nature of the proposed control structure offers many immediate benefits. Firstly, the MPC control can be implemented without replacing the existing feedback controller. Furthermore, it provides a clear framework to quantify the benefits in the use of advance real-time measurements over the nominal output-feedback strategy.