Thesis Summary 154

This section provides an outline of the research works as presented in the previous chapters.

The BEM method with wake induction correction models and stall correction models were examined through power performance analysis of the NREL/NASA Phase VI wind turbine. For wake induction correction, the Glauert model, the GH-Bladed model and the AeroDyn model demonstrate very similar results for the studied case. Without stall correction, the BEM method with purely 2D coefficients under-predicts power output from moderate wind speed to high wind speed. With V-C stall correction model, the combined coefficients provide improved power prediction. With D-S stall correction model, the BEM gets good results at low wind speeds and over-predicts power outputs at high wind speeds. A hybrid stall correction model was proposed and it shows better power prediction compared with the previous discussed models. It is therefore concluded that the accuracy of stall correction models are highly wind turbine dependent and operation condition dependent. Further validation of these models with more wind turbine measurements is needed.

The BEM blade design philosophy was investigated through two most typical small wind turbines: fixed-pitch variable-speed (FPVS) wind turbine and fixed-pitch fixed-speed (FPFS) wind turbine. The effects of the key rotor parameters on power

Conclusions and Future Work

and twist angle distributions are determinative to wind turbine performance. A blade design approach of searching optimal induction factors was developed in MATLAB code to obtain the optimal blade chord and twist angle distributions. The tip-hub loss and drag effect were included in the blade design of a 12kW FPFS wind turbine. Results show that the tip-hub loss and drag have apparent effects on both blade hub and tip region. Considering F (tip-hub loss factor) and drag effects, smaller blade chord and twist angle occur for Hub and tip region. This finding is particularly interesting for the blade tip and Hub design and power performance improvement. Three different linearisation strategies of blade chord and twist angle distributions were investigated. The un-linearised twist strategy (only chord is linearised) demonstrate higher power production compared with the linearised twist angle strategy (both twist angle and chord are linearised). Considering less materials and higher AEP, it is preferable to linearize chord according to the preliminary outer sections. A heuristic approach of blade design optimization through linearisation of radial profile of the chord and twist angle for FPFS small wind turbines was developed. This approach can be used in any practical FPFS wind turbine blade design and refurbishment.

The 2D CFD modelling and 3D CFD modelling were validated against measurements of the S809 airfoil and the NREL/NASA Phase VI wind turbine. Mesh dependency and turbulence dependency studies were conducted. In 2D CFD modelling, results show that the mesh node numbers around the airfoil affect the accuracy of the prediction. With a high mesh resolution, the accuracy can be improved but more computing time is needed. The SST transition model demonstrates better agreement in drag coefficient prediction than the fully turbulent SST k-ω model compared with measured results. The quasi-3D CFD modelling calculations produce very similar results in lift and drag coefficients prediction but consume more computing time compared with 2D CFD modelling. In 3D CFD modelling, a series of detailed flow characteristics were obtained including integrated forces and moments, blade surface pressure distributions and flow streamlines. Results show good qualitative and quantitative agreements with the measurements and other research works from literatures. The purposes of validation and deep insight view of detailed flows for stall phenomenon have been fully achieved. The comparative study of mesh and turbulence models is instructive for any kind of wind turbine CFD modelling and definitely represents a foundation for future work.

Conclusions and Future Work

The 2D CFD analysis and wind tunnel tests of the DU93-W-210 airfoil were implemented at relatively low Reynolds numbers from 2×105 to 5×105. The wind tunnel tests were conducted at three wind speeds of 10m/s, 15m/s and 25m/s in the Aerodynamics Laboratory at Hertfordshire University. The lift, drag and moment coefficients of the airfoil DU93-W-210 were firstly measured at this range of Reynolds numbers without any published data available. All the measured coefficients show the same trend at the three Reynolds numbers. The lift coefficients increase with Reynolds number and the drag coefficients decrease with Reynolds number, which verifies that a higher lift to drag ratio is expected at a higher Reynolds number. The stall angle moves from 14° to 12° with Reynolds number changing from 2×105 _{to 5×10}5_{, while the stall}

angle of the same airfoil is around 10° at Reynolds number of 1×106 _{[93]. The wind}

tunnel test results and the 2D CFD results show reasonable agreements. It is noted that the measured drag coefficients are higher than the CFD calculated drag coefficients. The discrepancies in drag coefficients are mainly due to the complex flows at the ends of the airfoil section, which were caused by the gap between the two ends of the airfoil section model and the wind tunnel side walls.

3D CFD analysis was performed for the two BEM-designed wind turbines. In the 3D CFD modelling of the FPVS wind turbine rotor, a series of calculations were carried out by fixing the tip speed ratio. The power performance of the rotor is well-predicted compared with the BEM methods. In order to have a further insight of the flow details, more calculations were done with a fixed rotor speed. The 3D CFD predicted blade surface streamlines demonstrate that before stall the flow direction is parallel to the chord-wise direction for the mix airfoil blade. It is also notice that the span-wise flow exists at high wind speeds. For the FPFS wind turbine rotor, 3D CFD calculations were performed at four wind speeds before and after stall. The calculated results were then compared with the BEM results. Good coincidences occur at 7m/s, 8.4m/s and 10m/s. The 3D CFD predicts slightly higher power output at high wind speeds compared with the BEM method using coefficients obtained from TUDelft wind tunnel test and the standard flat plate method. And the 3D CFD under-predicts power output compared with the coefficients extrapolated from modified flat plate method. It is verified that the CFD approach is able to provide a more detailed qualitative and quantitative analysis for wind turbine airfoils and rotors. With more advanced turbulence model and more

Conclusions and Future Work

considering 3D flow effects.