Escolar Documentos
Profissional Documentos
Cultura Documentos
art ic l e i nf o a b s t r a c t
Article history: The paper demonstrates the potential of an unsteady RANS 3D approach to predict the effects of skewed
Received 24 May 2015 winds on the performance of an H-type vertical-axis wind turbine (VAWT). The approach is validated
Received in revised form through a comparison between numerical and experimental results for a full-scale Darrieus turbine,
12 September 2015
demonstrating an improved prediction ability of 3D CFD with respect to both 2D CFD and semi-empirical
Accepted 20 September 2015
models based on the double multiple stream tubes method. A 3D URANS approach is then adopted to
Available online 22 October 2015
investigate the power increase observed for a straight-bladed small-scale turbine in a wind tunnel when
Keywords: the rotational axis is inclined from 0 to 15 from the vertical. The main advantage of this approach is a
Vertical axis wind turbine more realistic description of complex three-dimensional ow characteristics, such as dynamic stall, and
Offshore
the opportunity to derive local blade ow conditions on any blade portion during upwind and downwind
Skewed ow
paths. Consequently, in addition to deriving the turbine overall performance in terms of power coef-
Dynamic stall
CFD cient, a better insight into the temporal and spatial evolution of the physical mechanisms is obtained. Our
principal nding is that the power gain in skewed ows is obtained during the downwind phase of the
revolution as the end part of the blade is less disturbed by the wake generated during the upwind phase.
& 2015 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jweia.2015.09.010
0167-6105/& 2015 Elsevier Ltd. All rights reserved.
78 A. Orlandi et al. / J. Wind Eng. Ind. Aerodyn. 147 (2015) 7784
Experimental investigations on H-VAWT turbines have demonstrated elements. The external dimensions are 14D 30D 12D and the
an enhanced performance for relatively small tilt angles (up to 2530) wall distance is 1n10 4c, necessary for an averaged y o 5, fol-
depending on the shape of the rotor, whereas a reduced performance lowing a mesh sensitivity analysis performed in order to verify if
is produced by higher skew values (Mertens et al., 2003). This benet the grids adopted were sufciently ne to resolve the primary
might be explained by the fact that VAWT blades sweep out a ow features. Fig. 2 shows the analysis for the 2D model for the
cylindrical surface, as opposed to a planar surface for HAWTs. As a 17 m Darrieus VAWT and the 3D model for the small H-Darrieus
consequence, during misaligned ow operations, the swept area of the VAWT; the 3D model of the SANDIA turbine has been excluded
turbine is increased. from this analysis due to the high number of cells of the domain,
To account for skewed ow effects on VAWT performance, the but the mesh adopted is similar to that used by Howell et al.
efforts of researchers have focused on the implementation of (2010) and Zhang et al. (2013).
semi-empirical corrections in blade element momentum models
(Bianchini et al., 2012; Madsen et al., 2014). To the authors'
2.1. 2D and 3D mesh sensitivity analyses
knowledge, there are no studies in the literature presenting a 3D
CFD (Computational Fluid Dynamics) study on inclined VAWT
For all the cases presented, both for the 2D and the 3D models,
performance. The present study is aimed at contributing to a
mesh sensitivity analyses have been conducted, to assess the
better understanding of the physical processes that result in the
adequate mesh element number able to resolve the ow. As
measured performance enhancement reported for VAWTs in
regards the 2D model, three meshes were analysed (coarse,
skewed ows.
medium, ne) as shown in Table 1, keeping the wall distance xed
at 1.6n10 5 with y o 1.
The 3D mesh independence study was performed only for the
2. Numerical approach
skewed ow case. The same philosophy was then applied to the
SANDIA case. For the 3D model of the turbine in skewed ow, two
The two wind turbine congurations analysed in this work are:
grids have been considered (4 and 5), as shown in Table 2. In
general a larger number cases are analysed in order to conduct a
the SANDIA 17 m-diameter Darrieus-Type VAWT used to
sensitivity analysis, but in this case we considered the previous
validate the CFD modelling approach that has been adopted. It
work done in this area (see cited papers). All the 3D meshes were
has an aspect ratio (height to diameter ratio) equal to 1.02, two
blades with a NACA 0015 aerofoil section and a constant chord built considering a wall distance equal to 1.9n10 4c, resulting in
of 0.612 m. Rotational speed is xed at 38.7 RPM (rotational y o 5.
speed, in revolutions per minute) for aerodynamic force mea- The results of these analyses are shown in Fig. 1, where the Ct
surements and 42.2 RPM for cp measurements (Akins, 1989); (tangential force coefcient) values are compared.
a two bladed H-Darrieus VAWT with NACA 0018 aerofoil As regard the 2D analysis, the difference between mesh 2 and
sections and constant 0.08 m chord. The rotor height is 0.5 m mesh 3 is minor, and mesh 2 was adopted, in order to save time
and diameter is 0.755 m. This conguration was used in wind and computational resources without substantially affecting
tunnel tests to study the inuence of skewed ows (Mertens, accuracy. The same conclusion can be derived from the second
2002) allowing direct comparison with CFD predictions. graphs, showing a small gap between the two grids. So that MESH
4 was used for the 3D analysis. Furthermore, this meshing philo-
For all the simulations, two different grid levels have been sophy was subsequently adopted for the SANDIA turbine case.
adopted: a xed sub-grid with the external dimensions of the ow
domain, and a dynamic sub-grid that includes the VAWT geometry 2.2. CFD solver approach
and allows a relative motion with respect to the xed grid. This
grid arrangement utilizes the sliding mesh technique (Ansys, The CFD software used for the present analyses is FLUENT v.15,
2009) and allows the simulation of the rotating motion of the developed by Ansys Inc. This study is based on the URANS implicit
wind turbine with a steady RANS (Reynolds-averaged Navier- model, and turbulence is modelled using the k SST (Shear Stress
Stokes equations) or URANS (unsteady RANS) analysis. Transport) model. This turbulence scheme was adopted because of
The CFD process was validated for the SANDIA National Labs its aptitude in cases involving high adverse pressure gradients and
conguration using both 2D and 3D grids. The 2D mesh adopted is therefore smooth surface separations (Menter, 1993). The air was
a hybrid structuredunstructured mesh with around 460 000 considered as incompressible since the different cases studied did
elements. The ow domain external dimensions were 37D x 25D, not exceed a local Mach number greater than 0.3.
and the wall distance from the rst layer of cells is set at The setup settings for the simulations are shown in Table 3. The
1.6n10 5c with y (dimensionless wall distance) o1, where D is simulations are done using a First Order Implicit scheme for all the
the diameter of the turbine and c is the blade chord length. The 3D variables of the spatial discretization in the rst revolutions, after
grid contains around 2 800 000 elements with ow domain that increase the order of the schemes considered to prevent the
external dimensions of 15D 5D 4D and a wall distance equal to model from numerical diffusion errors. The convergence criteria
1.9n10 4c, resulting in y o 5. were set at 1n10 5 for all residuals.
For the analysis of the rotor in a skewed ow, only a 3D
approach was considered and the mesh contains around 1 800 000 Table 2
Details of the 3D grids for H-Darrieus turbine.
Table 1
Details of the 2D grids. Mesh No. of elements Nodes around Nodes Wall distance
blades spanwise
Mesh No. of nodes on blade No. of elements Wall distance
4 1 724 930 84 200 1.6 E 04
1 620 301 163 1.6 E 05 chords
2 1240 464 845 1.6 E 05 5 2 672 993 186 300 1.6 E 04
3 1920 700 292 1.6 E 05 chords
A. Orlandi et al. / J. Wind Eng. Ind. Aerodyn. 147 (2015) 7784 79
Fig. 1. Mesh sensitivity analysis for 2D model of 17-m Darrieus-Type VAWT (a) and for the 3D model of H-Darrieus VAWT (b).
Fig. 2. (a) External domain, (b) details of 3D grid for SANDIA 17 m Darrieus-Type wind turbine, and (c) of the H-Darrieus Type Wind Turbine, (d) volume mesh.
3. Results Table 3
CFD simulations: setup settings.
3.1. Verication of the CFD model
Solver
This paper focuses on two wind speed operating conditions: Type Pressure-based
TSR 4.6 (Tip Speed Ratio) (low-wind speed case) and TSR 2.33 Time Transient
(high wind speed case), for the SANDIA 17 m-diameter test case. Solution methods
Since chordwise pressures for this case were measured at the Pressure-velocity coupling PISO
equatorial station of the turbine blade, it is expected that the local
Spatial discretization
ow at this station is 2D and therefore that 2D CFD simulations Gradient Green-Gauss node based
give a good approximation of the ow. Normal and tangential Pressure Standard
Momentum Second order upwind
force coefcients were obtained by integrating the chordwise
Turbulent kinetic energy Second order upwind
pressure distributions as reported by Akins (1989). The normal and Specic dissipation rate Second order upwind
tangential force coefcients cn and ct are dened as Modied turbulent viscosity Second order upwind
Transient formulation
Fn
cn 1 Second order implicit
1=2 U 21 c
80 A. Orlandi et al. / J. Wind Eng. Ind. Aerodyn. 147 (2015) 7784
Fig. 3. Normal force coefcient and tangential force coefcient, calculated at TSR 4.6 (a, b) and at TSR 2.33 (c, d).
Fig. 7. cp ratio and TSR ratio for H-Darrius Turbine at different tilt angles. The ratios refer to the maximum values of the variables reached at different tilt angles normalized
by the maximum value calculated at 0 tilt angle. From wind tunnel test by Mertens (2002).
82 A. Orlandi et al. / J. Wind Eng. Ind. Aerodyn. 147 (2015) 7784
Fig. 8. Normal force and tangential force coefcient as a function of the azimuthal angle, at different tilt angles.
Fig. 9. Pressure eld calculated at 01 (left) and 151 (right), considering a blade section at 0.25 m distant from the tip in the less inuenced zone.
Darrieus turbine. Cp is dened as All results refers to the seventh revolution, where the differ-
ence between the average total torque coefcient Cm and the one
Power
Cp 3 of the previous revolution is within 1%. SANDIA turbine has been
1=2U 31 DH analysed at two different wind conditions:
where H is the blade height, m. Due to the better prediction of the
TSR 4.6 (corresponding to a free stream velocity of 7.345 m/s)
tangential force coefcient at higher wind speeds, the 3D CFD
with a time scale corresponding to 1/time-step (0.0043 s).
results are in closer agreement with experimental data and are
TSR 2.33 (corresponding to a free stream velocity of 14.5 m/s)
generally within 10%. The 2D CFD model over-predicts power at
with different time scales corresponding to 1/time-step
high wind speeds, as would be expected from the previous ana-
(0.00431 s), 0.5/time-step (0.00216 s) and 0.25/time-step
lysis. Fig. 4 also compares the data with those predicted using BEM
(0.00108 s).
(Blade Element-Momentum) approaches by Allet and Paraschivoiu
(1995) as well as another CFD approach (Torresi et al., 2013). The
As regards the simulations of the H-Darrieus turbine, the free-
advantage of the CFD approach becomes clearer at high wind- stream velocity was xed at 7 m/s and all the results refer to a
speed conditions, where greater accuracy in predicted power is time scale equal to 1/time-step at each TSR considered.
achieved. Fig. 6 shows good agreement of the k SST model with wind
tunnel test data, though the k RNG scheme gives poor results,
3.2. Rotor in skewed ow particularly at low wind speeds. The difference between the k
SST model and the measured data is within 10%, except at the
Since the purpose of this work is to determine the inuence of highest wind speed. This difference between these two models can
changes to the inclination of a VAWTs' rotational axis relative to be attributed to an overproduction in wall shear stresses by the
the ow direction on the rotor performances, a 3D CFD model is fully turbulent k RNG model, which led to a smaller averaged
necessary. The two bladed H-Darrieus VAWT conguration was torque value. In fact, as shown in Fig. 5, k RNG model gives
used, with measured wind tunnel test data available for a straight higher values of wall shear stresses than the one calculated by the
and inclined rotor (Mertens, 2002). A preliminary analysis in order k SST model. These generate forces which oppose the motion of
to assess the best turbulence model scheme was performed for revolution of the turbine, resulting a lower averaged torque coef-
this wind turbine, without skewed inow. This analysis considered cient and thus a lower power coefcient.
two of the most referenced turbulence models for cases including Both test cases have demonstrated that the k SST model can
strong adverse pressure gradients: k RNG (Re-Normalization simulate the power production of a VAWT with reasonable accu-
Group) and k SST. The comparison was based on the calculation racy. Consequently, the k SST turbulence model was adopted for
of the power coefcient Cp at different values of TSR. quasi-static simulations of the rotating H-VAWT conguration at
A. Orlandi et al. / J. Wind Eng. Ind. Aerodyn. 147 (2015) 7784 83
xed inclination or tilt angles. The inclination angle was varied previous blade on the oncoming one. The comparison also reveals
in 5 increments from 5 to 15. Higher inclination angles were not that the Mertens' model does not agree very well with the
considered since it is assumed that the H-VAWT is installed on an experimental data. The main reason is due to the DMST (Double-
offshore oating platform and these structures are designed not to Multiple Stream Tube) model not including the semi-empirical
exceed a limit of 715 oscillations about the vertical axis (Collu corrections that account for the dynamic stall. Indeed, this con-
et al., 2014). sideration is linked to the results shown in Fig. 7b. It is evident
Fig. 7a refers to the ratio of the maximum value of cp at different that, at a higher TSR ratio, which implies higher rotational velocity
tilt angles normalized by the maximum value at 0 skew angle; of the wind turbine (since the inlet velocity is xed at 7 m/s), and
Fig. 7b refers to the ratio of TSR at which the maximum cp is reached at higher tilt angles, the cp ratio values are closer to the one
at different values of , normalized by the maximum value at 0. obtained in the wind tunnel tests corresponding to the same tilt
The cp ratio has been considered rather than the cp since only the angle. The combined effects of the increased rotational velocity
rst was available in the report considered. The CFD results are also and the augmentation of the tilt angle lead to a smaller inuence
compared with measured wind tunnel results and results from a of the dynamic stall and so a better matching between the Mer-
BEM model developed by Mertens (2002). The BEM model under- ten's model and experimental data. However, also shown in this
predicts the power in a skewed ow, while the CFD results are in graph, the present model matches the TSR ratio measurements
good agreement with the wind tunnel measurements. better than the Mertens' model.
This means that the present model can well describe the As seen from Fig. 7, the cp increases with due to the increases
behaviour of the ow acting on the blade sections, such as the rise of the blade tangential force coefcient. The normal and tangential
of dynamic stall and the effects of the wake generated by the force coefcients predicted at the equatorial plane of the rotor are
84 A. Orlandi et al. / J. Wind Eng. Ind. Aerodyn. 147 (2015) 7784
shown in Fig. 8a and b, respectively. As regards the normal forces, performed, with the aim to separate and analyse the causes that
small changes are seen in the rst half of the revolution, where the lead to an increase of the coefcient of power cp. The present work
maximum difference between the cases at 0 and 15 is offers a possible interpretation of this phenomenon, which was
equal to 1.8%, while downwind the blade experiences normal studied through experimental tests and only briey using a semi-
forces that increase with tilt angle. empirical model by Mertens. This study founds that the gain is due
Regarding the tangential force coefcient, which has a direct to the downwind part of the rotor being less disturbed by the
impact on the rotor torque and power, although slightly reduced wake generated by the blades in the upwind part. Since the
values are experienced over the upwind cycle (maximum of 4.5% interest in the future use of VAWTs is primarily on oating plat-
lower than the case at 0), signicant increases are experi- forms for offshore applications, it could be interesting to evaluate
enced over the downwind cycle. Consequently this increase in the the inuence of real motion on the rotor performances. However,
tangential force coefcient is the reason for the augmentation of it is important to note that it has been possible to determine and
the normalized power. In fact, Fig. 8 shows that the maximum localize the physical mechanisms described above using a rela-
value of cn and Ct is reached for the case at 15: tively simple numerical model, i.e. with only 2.5 millions of cells
cn;MAXf1801 o o 3601g; 151 ct;MAXf1801 o o 3601g; 151 for a full scale turbine and about 2 million for the small turbine, a
2:32 7:1 relatively simple turbulence model (two-equations model), and
cn;MAXf1801 o o 3601g; 01 ct;MAXf1801 o o 3601g; 01
appropriate time steps that were found as a good compromise
4
between computational costs and accuracy in results.
This benecial effect is because the bottom part of the rotor in
the downwind section is less inuenced by the passage of the
previous blade, i.e. less wake interference. This is shown in Fig. 10, References
showing colour shaded velocity contours on a plane parallel to the
ow direction. Akins, R.E., 1989. Measurement of Surface Pressure on an Operating Vertical-axis
During the revolution, the wind over the upwind cycle is Wind Turbine. Technical Report SAND89-7051, SANDIA National Labs.
decelerated by the passage of the blades, i.e. there is a momentum Allet, A., Paraschivoiu, I., 1995. Viscous ow and dynamic stall effects on vertical-
axis wind turbines. Int. J. Rotat. Mach. 2 (1), 114.
loss associated with the work extracted by the rotor. This wake Ansys, 2009. Ansys Fluent 12.0. User's Guide, Ansys Inc.
moves downstream in the wind direction creating a zone of low Bekiropoulos, D., Rie, R.M., Lutz, T., Krmer, E., Matha, D., Werner, M., Cheng, P.W.,
momentum ow that is crossed by the advancing blade. When the 2012. Simulation of unsteady aerodynamic effects on oating offshore wind
turbines. In: 11th German Wind Energy Conference, 78 November 2012,
rotor is not inclined to the ow, the entire downwind blade Bremen, Germany.
encounters the wake from the upwind blade. However, at different Bianchini, A., Ferrara, G., Ferrari, L., Magnani, S., 2012. An improved model for the
a lower part of the blade is outside of this wake and thus performance estimation of an H-Darrieus wind turbine in skewed ow. Wind
Eng. 36 (6), 667686.
experiences a higher dynamic pressure and an increase in local Borg, M., Collu, M., 2015. A comparison between the dynamics of horizontal and
aerodynamic forces. vertical axis offshore oating wind turbines. Philos. Trans. R. Soc. A: Math. Phys.
Similarly, Fig. 11 shows the effect of the wake from the passage Eng. Sci. 373 (2035).
Borg, M., Shires, A., Collu, M., 2014. Offshore oating vertical axis wind turbines,
of the preceding blade on the downwind cycle in terms of vorticity dynamics modelling state of the art. Part I: aerodynamics. Renew. Sustain.
contours, and highlights that at higher tilt angles the vorticity is Energy Rev. 39, 12141225.
reduced in this region. Collu, M., Borg, M., Shires, A., Rizzo, N.F. Lupi, E., 2014. FloVAWT: further progresses
on the development of a coupled model of dynamics for oating offshore
The static pressure eld is shown in Fig. 9 for a blade aerofoil
VAWTs. In: Proceedings of the ASME 2014 33rd International Conference on
section at 2401 and at a distance of 0.25 m from the lower Ocean, Offshore and Arctic Engineering, 813 June 2014, San Francisco, USA.
blade tip (i.e. far enough to be not inuenced by the tip vortices). Giorgetti, S., Pellegrini, G., Zanforlin, S., 2015. CFD investigation on the aerodynamic
This qualitative comparison shows some signicant differences in interferences between medium-solidity Darrieus vertical axis wind turbines.
Energy Procedia, in press.
the local pressure eld as the rotor is tilted to 151, especially Howell, R., Qin, N., Edwards, J., Durrani, N., 2010. Wind tunnel and numerical study
on the suction surface, resulting in a higher value of lift and an of a small vertical axis wind turbine. Renew. Energy 35, 412422.
augmentation in the power coefcient. Kinzel, M., Mulligan, Q., Dabiri, J.O., 2012. Energy exchange in an array of vertical
axis wind turbines. J. Turbul. 13 (38), 113.
Madsen, H.A., Barone, M., Roscher, B., Deglaire, P., Arduin, I., et al., 2014. Comparison
of aerodynamic models for vertical axis wind turbines. In: Journal of Physics:
4. Conclusions Conference Series, vol. 524, IOP Publishing, p. 012125, Copenhagen, Denmark.
Menter, F.R., 1993. Zonal two equation k-omega model for aerodynamic ows. In:
24th Fluid Dynamics Conference, 69 July, Orlando, Florida.
It has been shown that the CFD simulations using a URANS Mertens, S., Van Kuik, G., Van Bussel, G., 2003. Performance of an H-Darrieus in the
model are a promising way to investigate the complex phenomena skewed ow on a roof. J. Sol. Energy Eng. 125, 443540.
Mertens, S., 2002. Wind energy in urban areas: concentrator effects for wind tur-
that are present in an analysis of a VAWT. After initially validating
bines close to buildings. Refocus 3 (2), 113.
the CFD model, the second part of the paper utilizes the method to Shires, A., Design optimisation of an offshore vertical axis wind turbine. Proc. ICE-
consider the performance increments that have been observed in Energy 166 (1) (2013) 718.
Tongchitpakdee, C., Benjanirat, S., Sankar, L.N., 2005. Numerical simulation of the
skewed ow. The 3D CFD model shows better prediction of the
aerodynamics of horizontal axis wind turbines under yawed ow conditions. J.
behaviour of the uid, which involves unsteady three-dimensional Sol. Energy Eng. 127 (4), 464474.
effects such as dynamic stall. However, some differences from the Torresi, M., Fortunato, B., Camporeale, S.M., 2013. Modello CFD per il calcolo delle
experimental data have been displayed and it could be useful to prestazioni e degli effetti di scia di turbine eoliche ad asse verticale, La
Termotecnica.
investigate these further using more complex models, LES (Large Zhang, L.X., Liang, Y.B., Liu, X.H., Jiao, Q.F., Guo, J., 2013. Aerodynamic performance
Eddy Simulation) or DES (Detached Eddy Simulation) models for prediction of straight-bladed vertical axis wind turbine based on CFD. Adv.
example, in order to give a more precise description of these Mech. Eng., 5, 905379, http://dx.doi.org/10.1155/2013/905379.