Date of Award


Document Type

Doctoral Thesis

Degree Name

Doctor of Philosophy


Mechanical, Biomedical & Manufacturing Engineering

First Advisor

Dr. Andrew Cashman


Wind energy has witnessed a consistent expansion over the past decade, especially with the move to offshore installation. There is an increasing need to further exploit superior offshore wind resources, which is pushing multi-megawatt wind turbines into deeper water locations where the current popular horizontal axis wind turbine configuration is not entirely suitable. In particular, there has been a renewed interest in the vertical axis wind turbine (VAWT) configuration due to its inherent design attributes for an offshore floating application and also its potential to provide a significant reduction in the system cost of energy. However, challenges still remain as the current offshore VAWT technology status lags greatly behind its horizontal axis counterpart. This research concentrates on the aerodynamic design and simulation of a large-scale stall-regulated H-type VAWT for this offshore application. At this large-scale, the VAWT's blades will operate at high Reynolds numbers and encounter dynamic stall at low tip-speed ratios (TSRs). A validated computational fluid dynamics (CFD) method was developed to stimulate the unsteady aerodynamics experienced by a VAWT at this scale. The performance of Unsteady Reynolds-Averaged Navier-Stokes (URANS) and Detached Eddy Simulation (DES) modelling methods are compared in simulating the aerodynamics of an isolated NACA 0018 blade experiencing Darrieus pitching motion. The URANS turbulence models employed were the Spalart-Allmaras (S-A) model and the k-w SST model. Investigations were conducted to ensure satisfactory independency of the solution for both spatial and temporal discretisations, respectively. A quantitative assessment identified the S-A model as the most applicable for a VAWT design study, as it showed the most desirable compromise between model fidelity and computational requirement. A qualitative analysis revealed that the thick VAWT blade creates a dynamic stall vortex topology highly concentrated at the trailing edge region. Increasing the Reynolds number showed to be beneficial to the blade's aerodynamic performance as a higher maximum tangential force coefficient is attained, owing to the delay in flow separation to much higher angles of attack. Increasing the freestream turbulence intensity similarly delayed the dynamic stall onset and the blade flow reattachment feature following the dynamic stall event. Investigation of the blade mounting point position during dynamic stall showed the chordwise range x/c=0.2-0.3 resulted in the lowest blade pitching moments. A low-order model (LOM) was developed to provide a rapid calculation of the VAWT performance and improve the design process efficiency. The LOM incorporates different sub-models for various aerodynamic effects, including a Beddoes-Leishman (B-L) dynamic stall model to account for unsteady dynamic stall effects. To provide enhanced numerical efficiency and stability, an iterative time-advancement scheme with adaptive under-relaxation has been integrated into the developed LOM. A comparative study of the LOM an the CFD model was undertaken to assess predictive accuracy with actual VAWT aerodynamic blade force experimental data and power coefficient measurements. The LOM showed good agreement with the CFD model and the measurements with a low computational cost requirement. The CFD results identified that as the TSR was increased, the rotating tower downwind wake region became increasingly more skewed and more influential over a wider range of downwind azimuthal angles. The 2D CFD model captured the qualitative shape of the VAWT performance curve but greatly overestimated efficiency at all the simulated TSRs. An approach for computing the B-L dynamic stall model steady and unsteady airfoil parameters using CFD was investigated to extend its applicability for VAWTs. This method permits the calculation of the blade dynamic stall characteristics over a range of reduced pitch rate by employing a user-defined sliding mesh motion. This technique was shown to be successful and can be employed where the required B-L model input empirical coefficients are not readily available and particularly useful for new airfoils. The variation in the blade Reynolds number over the VAWT operating envelope is also considered by this approach. The geometrical and operating specifications for a variable-speed 5 MW VAWT were identified. The VAWT solidity, blade orientation, blade aspect ratio and the support strut design was investigated. A VAWT solidity of 0.263 maximised the aerodynamic efficiency and ensured blade dynamic stall was avoided at the optimum TSR function regime. Results showed the concave-out configuration for the cambered blade increased the peak torque coefficient by 4.5% compared to the concave-in arrangement. It was observed that the blade aerodynamic forces were more sensitive to the blade orientation at low TSRs than at high TSRs. A non-prismatic tapered strut design was utilised and created a 6.5% reduction in the peak efficiency. Structural analysis of the blade structure subject to a critical load case was investigated with two methods, an analytical model and a finite element (FE) model. It was shown that the utilisation of a composite blade topology can resist the induced flapwise loading and the material strains were contained within their allowable limits. The analytical approach was demonstrated to be a quick and accurate technique to compute the composite blade strain distribution when compared to the FE model results. A 3D CFD model was employed to examine the 5MW VAWT aerodynamic phenomena and wake evolution. Dynamic stall causes the VAWT streamwise wake to become increasingly asymmetrical as the TSR is reduced. The impact of the blade tip vortex varies wit the azimuthal angle and the upwinde tip vortex is more intense compared to the downwide tip vortex. Blade end plates were investigated displaying a 4.71% increase and a 23.1% decrease in the mean torque coefficient for the upwind and downwind phases, respectively. A 0.73% reduction in peak efficiency was realised by utilising blade end plates. A tower fairing was also examined and was demonstrated to be an effective device to eliminate the vortex shedding created by the rotating VAWT tower.


I would like to acknowledge the DJEI/DES/SFI/HEA Irish Centre for High-End Computing (ICHEC) for the provision of computational facilities and support. The following high-performance computing National Service projects were utilised in this work: cieng003c, cieng006c, cieng007c, cieng008c and cieng009c. Important elements of this research would not have been possible without this extremely valuable resource and I am truly grateful to have had the opportunity to use the Fionn supercomputer.

I would like to thank the University of Glasgow for the provision of experimental data and literature concerning dynamic stall research. As well, I want to acknowledge the National Research Council of Canada Low Speed Aerodynamics Laboratory for permitting experimental data to be used in this research.

Access Level