Sei sulla pagina 1di 15

Original Article

Proc IMechE Part A:


J Power and Energy
A numerical study of blade thickness 226(7) 867–881
! IMechE 2012

and camber effects on vertical axis Reprints and permissions:


sagepub.co.uk/journalsPermissions.nav

wind turbines DOI: 10.1177/0957650912454403


pia.sagepub.com

Louis Angelo Danao1,2, Ning Qin1 and Robert Howell1

Abstract
This article presents the results of a two-dimensional computational study into the effects of rotor blade thickness
and camber on the performance of a 5 kW scale vertical axis wind turbine. Validation is provided by reference to
experimental data for a pitching aerofoil with dynamic stall phenomenon. The performance of the turbine is mapped out
for a variety of different tip speed ratios and detailed investigations are presented to determine how and, most import-
antly, why the turbine performance varies with thickness and camber as it does. The rotor blades chosen were the
NACA0012, NACA0022, NACA5522 and LS0421. The turbine rotor has a diameter of d ¼ 3.1 m with a blade chord
length of c ¼ 0.18 m. Over the range of tip speed ratios examined, the NACA0012 profile performed with the highest
overall performance of 50% at  ¼ 3.5. Slightly cambered aerofoils (such as the LS0421) can improve the overall
performance of the vertical axis wind turbine, whereas a camber of 5% (NACA5522) results in unfavourable perform-
ance. Of the cambered blades tested, the LS0421 performs the best with maximum CP of 0.40 at  ¼ 3.5. A camber along
the blade path causes the blades to generate higher values of torque in both the upwind and the downwind regions.
It was also determined that inverted cambered profiles produce power mostly in the upwind region.

Keywords
Vertical axis wind turbine, computational fluid dynamics, unsteady, thickness, camber

Date received: 1 February 2012; accepted: 11 June 2012

Introduction
having to yaw into the wind and lower production
Wind energy is becoming an increasingly important costs as the blades can be of constant profile along the
sustainable energy source. This is mainly due to span (H-type). While horizontal axis machines
changes in national and international energy policies are highly developed, the understanding of VAWTs is
in response to climate change. Many of the large-scale lacking. Much of the research on VAWT design was
commercial wind farms have been built in the UK carried out as long ago as the late 1970s and early
(including the largest off-shore wind farm) and electri- 1980s, notably at the US Department of Energy
city generation from wind now exceeds 5 GW installed Sandia National Laboratories2–5 and in the UK by
capacity with a further 9.5 GW in application. The UK Reading University, and Sir Robert McAlpine and
Government had ambitious targets under the Kyoto
Protocol for renewable energy generation of 10% of
UK consumption by 2010. Although this target was 1
Department of Mechanical Engineering, University of Sheffield, UK
2
not achieved, 2.5% of the UK’s power generation Department of Mechanical Engineering, University of the Philippines,
came from wind energy in 2009.1 Philippines
The vast majority of installed wind power solutions
Corresponding author:
are horizontal axis wind turbines (HAWTs). However, Louis Angelo Danao, Department of Mechanical Engineering, University
there may be advantages for using vertical axis wind of Sheffield, Sheffield S1 3JD, UK.
turbine (VAWT) machines which include them not Email: louis.danao@sheffield.ac.uk
868 Proc IMechE Part A: J Power and Energy 226(7)

Sons Ltd (through their subsidy VAWT Ltd) who stall, which is also the key aerodynamic phenomenon for
erected several prototypes including a 500 kW version VAWTs relating to performance.
at Carmarthen Bay.6 However, when it became accepted
that HAWTs were more efficient at these large scales,
VAWT kinematics
interest in VAWT designs waned and HAWTs have
since dominated the market. Little research can be One blade of the VAWT is shown in Figure 1(a). As the
found in the last couple of decades on VAWTs, but VAWT rotates with an angular velocity
in a flow of
analysis tools have improved immeasurably since then. wind speed U1, the velocity of the wind relative to the
Fundamental studies relating flow physics and perform- blade changes and is given by
ance are lacking in the literature and significant work is
required to determine the true potential of VAWTs. * * *
W ¼ U1 þ V ð1Þ
For example, it has never been conclusively shown
that HAWTs are fundamentally more aerodynamically *
efficient than VAWTs. It is often stated that there are where V ¼ 
r and r is the radius of the VAWT.
some advantages of VAWTs over HAWTs but again, This velocity fluctuates from a maximum of
nothing has ever been proven. It has been suggested ( þ 1)U1 to a minimum of (  1)U1, where  is the
that VAWTs may be more appropriate than HAWTs tip speed ratio. At the same time, the angle of attack 
at very large scale (10 MWþ) due to the alternating also varies periodically between the positive and nega-
gravitational loading on a HAWT blade becoming tive values. The magnitudes of the relative velocity and
excessive. the angle of attack are given by
Considerable improvements in the understanding
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
of VAWTs can be achieved through the use of W ¼ U1 1 þ 2 cos  þ 2 ð2Þ
computational fluid dynamics (CFD) and experimental
 
measurements. However, in many circumstances, 1 sin 
experimental data are not available due to the cost and  ¼ tan ð3Þ
 þ cos 
complexity of creating different models for testing. For
this reason, computational modelling is used in this art- where  is the azimuth angle and is measured from the
icle to show how the performance of VAWTs changes vertical Y-axis in the anti-clockwise direction.
with the thickness of rotor blade profiles and camber. An interesting feature in the variation of the angle of
In order to validate the CFD study, a valuable set of attack is that it closely resembles a sine wave, as shown
pitching aerofoil experimental data were used to com- in Figure 1(b). This perceived variation is relative to a
pare various turbulence models for flow with dynamic reference frame attached to the rotating VAWT with its

Figure 1. VAWT fundamentals: (a) forces and velocities on a blade; (b) angle of attack  variation as a function of azimuth angle  at
 ¼ 2.
Danao et al. 869

origin at the VAWT axis. As the tip speed ratio  boundaries are equivalent to 5d from the VAWT
increases, the skewness of the  variation becomes axis while the downwind boundary is 10d away.
less and the profile comes closer to the sine wave The domain is composed of one stationary, outer
(zero skew). sub-domain and one rotating, inner sub-domain. The
interface between the two sub-domains is 5c from
the VAWT blades.
The mesh around the blade is made up of a struc-
Simulation methodology tured boundary layer mesh and a fully unstructured
mesh beyond (Figure 2(b)). The height of the first cell
Model geometry is 2  105 m which gives an average of the yþ values
The two-dimensional (2D) VAWT used in this study is from the flow solutions of around 1 with a maximum
a representation of a typical 5 kW VAWT available in not exceeding 5. Equally spaced grid points are placed
the market. This VAWT model is a three straight-blade around the aerofoil and a structured boundary layer
H-type model, similar to the previous studies at the mesh with 25 layers (about 1%c) is inflated from the
University of Sheffield7,8 The computational domain blade surface with a growth rate of 1.05. The fully
(Figure 2(a)) is such that the boundaries are sufficiently unstructured mesh is grown from the structured bound-
far from the rotor to minimise boundary effects such as ary layer mesh with a growth rate of 1.1 (Figure 2(c))
insufficient wake development and blockage conditions. and a maximum edge length of ½c in the rotating sub-
The rotor has a diameter of d ¼ 3.1 m with a blade domain. The total number of cells for the entire model
chord length of c ¼ 0.18 m. The upwind and side is approximately 230,000, including the rotating zone

Figure 2. Details of the 2D CFD model for a single VAWT blade: (a) computational domain; (b) leading edge mesh; (c) blade mesh;
and (d) periodic convergence of moment coefficient.
870 Proc IMechE Part A: J Power and Energy 226(7)

around the blades and the far-field stationary domain. sensitivity analysis to be performed. Both time-step
The mesh resolution is much finer than previous work size and convergence criteria were successively refined
done at the University of Sheffield7,8 in order to better independently until negligible changes in torque coeffi-
resolve the dynamically stalled flow around the blade cient were observed.
aerofoil.
Grid independence was checked by successively
refining surface parallel mesh dimensions until
Blade geometry
negligible changes in the torque coefficient were Symmetric profiles used in the simulations were the
observed. It has been determined that 1500 equally NACA0012 and NACA0022 while cambered profiles
spaced nodes on the blade surface is the level at used were NACA5522 and LS0421 (NACA, National
which grid independence is attained. This is comparable Advisory Committee on Aeronautics; LS, low-speed
to the density of nodes from Consul et al.9 and is coar- aerofoil series). Both positive and negative cambers
ser than that of Ferreira et al.10 It is fair to comment were considered and the blade profiles can be seen in
that having 1500 points is somewhat large. However, Figure 3.
the use of equal spacing leads to such a high number to
satisfy density requirements in the leading edge and the
trailing edge areas. Maximum edge length for the
unstructured grid was also checked and the most
Validation of turbulence model
reasonably accurate setting was used. Growth rates To properly select the appropriate turbulence model for
for both structured boundary layer and unstructured the problem, a pitching aerofoil study was conducted
mesh were kept constant throughout. for validation. Experiments done by Lee and
Gerontakos11 on a pitching NACA0012 provided the
dynamic data of the lift, drag and moment coefficients,
Numerical set-up and streamline plots required for both force and
Ansys Fluent 6.3 was used for the numerical computa- flow validations. The blade has a chord length of
tions of the study. Simulations conducted represent flow c ¼ 0.15 m and executes a sinusoidal pitching motion
conditions of a constant inflow of 5 m/s with free-stream of about 0.25c with a mean angle of attack of 10 , a
turbulence intensity Tu ¼ 0.1% and free-stream turbu- pitching magnitude of 15 and a reduced frequency
lence length scale l ¼ 0.01 m. In Fluent, a fully unsteady, of  ¼ 0.1. The free-stream wind velocity is 13 m/s
segregated solver was selected and all spatial discretisa- with turbulence intensity of 0.08% corresponding to a
tion terms were set to second order to accurately capture Reynolds number based on chord of around
the unsteady flow behaviour. The numerical time-step Rec ¼ 1.35  105.
size was set to an equivalent of ½ rotation of the It should be noted that the pitching motion does not
VAWT. Simulations were allowed to run until periodic fully capture the actual flow around a VAWT blade
convergence was observed in the predicted moment coef- since there is also a varying incoming velocity magni-
ficients. This was seen to take place after five full rota- tude aspect around a VAWT blade that is not present
tions of the VAWT (Figure 2(d)). in an aerofoil pitching about a fixed point. The rates of
Time-step convergence was determined when the change in the angle of attack are also different which
residuals fell below 1  104 and a residual drop of at can lead to a different flow behaviour between the two.
least two orders of magnitude in each time step for Regardless of this, it is believed that the dynamic inter-
all conserved variables. Verification of time-step size actions of a pitching aerofoil with a moving fluid are
and convergence criteria acceptability necessitated a the closest possible validation case available (due to the

Figure 3. Blade profiles used in the study: (a) NACA0012; (b) NACA0022; (c) NACA5522; (d) inverted NACA5522; (e) LS0421; and
(f) inverted LS0421.
Danao et al. 871

lack of detailed VAWT experimental data) with the


flow closely matching that of a moving VAWT blade
Force predictions
in as far as lift and stall are concerned. This is by far a Figure 4(a) and (b) shows the lift coefficient loops
better method compared to static aerofoil validation versus the angle of attack for different turbulence
because of the dynamic stall being similar. Another models. The static and dynamic experimental results
point that supports the use of this method is that the are plotted for comparison. It can be observed that in
variation of the angle of attack, as seen by a VAWT the upstroke, almost all the turbulence models accur-
blade without velocity induction, closely resembles a ately predict lift with the exception of the LR SST k–!
sine wave albeit skewed. The authors believe that this model. There is some oscillation in the lift that the LR
similarity would result in similar flow behaviour and SST k–! model predicts that is seen as the aerofoil
dynamic stall characteristics in the sense of testing tur- pitches upward, which is not present in the experimen-
bulence modelling. tal results. Maximum lift is not captured by the RNG
A fully structured O-grid was used for the simula- k–" and LR SST k–! models as they prematurely lose
tions. The aerofoil grid comprises 1500 nodes over the lift and proceed to drop even before reaching the max-
surface with a first cell height of 2  105 m ensuring a imum angle of attack. It is necessary to capture the
yþ & 1. The cells expand from the wall at a growth rate maximum lift because this has a significant effect on
of 1.1. The boundary of the domain was set to 20c from the overall performance of the VAWT.
the aerofoil. The total model size is approximately In the downstroke, all models under predict the lift
275,000 cells. The pitching motion was controlled by with the SST k–! model being the closest to the experi-
means of a user-defined function that prescribes the ment results and the S–A being the farthest away.
angular velocity of the entire domain to match the Overall, the most accurate model is the SST k–!
angular position dictated by the sine wave function model as far as lift is concerned. As mentioned previ-
 ¼ O þ A sin(!t), where O ¼ 10 , A ¼ 15 and ously, the LR RNG k–" model was also tested but
! ¼ 17.33 rad/s. A time-step size of t ¼ 0.001154 s results are not presented because they almost exactly
(equivalent to 0.1c/U1) was used. match the results of the fully turbulent RNG k–"
Different models were tested to find the most appro- model.
priate turbulence model for this study. The one-equation Figure 4(c) and (d) shows the drag and moment
Spalart–Allmaras (S–A) model, and the two-equation coefficient loops for the SST k–!. The drag is slightly
renormalisation group (RNG) k–" and shear stress over predicted by the SST k–! model at  ¼ 25 versus
transport (SST) k–! models were selected for the experiment results. The pitching moment is closely
study, all with low Reynolds (LR) number capabilities. calculated by the SST k–! model with an exception in
The S–A model automatically detects LR option if the the region close to the maximum angle of attack as the
mesh is resolved fine enough down to the wall.12 For the computed maximum moment coefficient is more than
RNG k–" and SST k–! models, LR effects through vis- double that of the experiment results. Again, the SST
cous damping are enabled to account for LR modelling k–! model is the best candidate for the turbulence
in the near-wall region. Runs were also conducted with- model with the best agreement to experiment results
out enabling the LR options for models RNG k–" and for this pitching aerofoil case.
SST k–! to check if not having LR capabilities can
impose significant differences on the predicted flow fea-
tures and force estimates.
Flow predictions
For this validation study, the initial mesh was con- The good agreement in the force predictions of the SST
structed following the description mentioned earlier. In k–! model necessitates further investigation. The flow
order to confirm grid independence, a second and third physics was checked to establish its suitability to
mesh were constructed with twice and half the resolution VAWT simulation. Boundary layer snapshots of the
in the wall and wall-normal directions. It was deter- pitching aerofoil at different stages are presented in
mined that the initial mesh was independent as the Figure 5. Both numerical results from this study and
results compared to the finer mesh showed negligible the corresponding experimental results from Lee and
difference in lift coefficient predictions. However, when Gerontakos11 are shown. At angles of attack of less
compared to the coarser mesh, significant differences in than 11 (not shown) in the upstroke, flow is still
the results were observed. All cases were run until full fully attached. Afterwards, flow reversal starts to
periodic convergence and it was observed that this hap- spread from the trailing edge upward. It is not apparent
pened after just two oscillation cycles. Time-step conver- in both Figure 5(a) and (f) but a closer inspection at the
gence was based on a residual drop to 1  106 and a trailing edge for  ¼ 16.75 " indicates reversed
drop of at least three orders of magnitude in each time flow already creeping up towards the leading edge
step. (Figure 5(k)). This is consistent with observations
872 Proc IMechE Part A: J Power and Energy 226(7)

from the experiment as mentioned in Lee and


Gerontakos.11 After the flow reversal phenomenon
Results
that precedes dynamic stall, the formation of a separ- After the validation of the turbulence models, a wide
ation bubble in the leading edge is observed (Figure range of  was tested for all blade profiles to capture
5(b)). This separation bubble rapidly grows and even- the power coefficient (CP) curves that depict a good
tually evolves into the dynamic stall vortex that is con- picture of VAWT performance. Figure 6(a) shows the
vected downstream (Figure 5(c) and (d)) until it is CP curves for the symmetric aerofoil cases. The
detached from the aerofoil surface after which sepa- NACA0012 performs best at  ¼ 3–4 with maximum
rated flow (Figure 5(e)) is observed to persist until CP of 0.50 at  ¼ 3.5. Figure 6(b) shows the moment
reattachment. The numerical results accurately capture coefficient plots of one blade for both profiles at
the development of the dynamic stall vortex, as can be  ¼ 3.5. Between the two, the thinner blade produces
seen in Figure 5(f) to (j). This is a strong indication that more torque in the upwind  ¼ 0 to 180 , while there
the SST k–! model is appropriate because the dynamic is not much difference in torque produced in the down-
stall phenomenon is properly captured. wind  ¼ 180 to 360 .
Based on the results obtained in this validation For all the blade profiles, maximum torque is
study, it is established that the SST k–! is the better achieved at close to  ¼ 90 with  ¼ 3.5. Also, the
turbulence model that satisfactorily predicts not only kink in the torque ripple at  ¼ 270 is indication of
the forces acting on the pitching aerofoil but also the the blade’s interaction with the wake of the VAWT
physics of flow that accompanies it. This verifies the post which slightly reduces the torque generated. The
validity on the use of the SST k–! model as an appro- general behaviour of the torque ripple as a blade goes
priate turbulence model for dynamic flow simulation one full rotation is that of the high torque generated in
that is required of an aerofoil in VAWT motion. This the upwind due to unperturbed, incident wind hitting
has been one of the most comprehensive validation the blade, and comparatively poor downwind perform-
studies done for a VAWT as both dynamic force and ance dictated by reduced incident wind velocity as well
flow predictions are compared against experiments. as blade–wake interactions.

Figure 4. Force coefficient loops for the pitching aerofoil: (a) lift coefficient loops for SST k–! and S–A; (b) lift coefficient loops for
RNG k–" and LR SST k–!; (c) drag coefficient loop for SST k–!; and (d) moment coefficient loop for SST k–!.
S–A: Spalart–Allmaras; RNG: renormalisation group; LR: low Reynolds; and SST: shear stress transport.
Danao et al. 873

symmetric NACA profiles were tested, namely


Stalling of symmetric aerofoils NACA0012 and NACA0022. Force coefficients on a
In order to examine the effects of blade thickness on the single blade of the VAWT were extracted from the simu-
stalling behaviour and performance of the VAWT, two lations to aid in the analysis. Instead of using standard

Figure 5. Visualisations for different angles of attack from Lee and Gerontakos:11 (a) 16.75 "; (b) 21.9 "; (c) 22.4 "; (d) 24.7 "; (e)
14.1 #; (f) to (j) are numerical results for SST k–! corresponding to a to e with contours of vorticity superimposed (arrows indicate
the direction of aerofoil travel). (k) flow reversal near trailing edge,  ¼ 16.75 ".
874 Proc IMechE Part A: J Power and Energy 226(7)

aerodynamic definitions of the different dimensionless This is consistent to the moment coefficient plot
coefficients, more appropriate expressions were adopted (Figure 6(b)) where the NACA0012 clearly operates
for lift, drag and moment coefficients and are as follows better even over the entire rotation versus the thicker
NACA0022. This contradicts the study performed by
L D M Healy13 where his model predicts thicker aerofoils to
Cl ¼ 1 2
, Cd ¼ 1 2 , Cm ¼ 1 2 ð4Þ perform better than thinner ones. But it is important
2 V c 2 V c 2 V cr to point out that Healy’s model is based on the multi-
ple-streamtube model that does not account for blade–
where V is the blade linear velocity as defined in the wake interactions in the downwind region and uses
‘VAWT kinematics’ section, c the blade chord and r the static aerofoil data from Jacobs and Sherman14 for per-
rotor radius. The use of blade linear velocity instead of formance prediction.
the relative velocity in the definition of the dynamic To understand why the thinner NACA0012 blade
pressure is intentional. For a VAWT blade, the relative performs better, the details of the performance of the
velocity is constantly changing in both magnitude and two blades as they go around one full rotation was
incidence. To facilitate the comparison of the force investigated. In the upwind, the NACA0012 generates
coefficients between different aerofoil profiles, the use the highest torque and reaches maximum at around
of a constant reference velocity is preferred. Between  ¼ 90 . This is true for both  ¼ 2.5 (Figure 6(d)) and
the aerofoils tested, the NACA0012 performs better  ¼ 3.5 (Figure 6(b)). Low  present highly unsteady
with a good power coefficient across a wide range of performance characteristics and flow features. In the
tip speed ratios. Figure 6(c) shows the pressure coeffi- second half of the upwind at  ¼ 2.5, the blade is stalled
cient on the blade surface for the two symmetric blades and vortices are constantly shed over time. As the blade
at  ¼ 90 and  ¼ 3.5. It can be seen that the passes downwind between  ¼ 180 and 240 , the torque
NACA0012 experiences greater pressure difference curve of the NACA0012 has a peak lower than that of
between the suction side and the pressure side of the the NACA0022 (Figure 6(b)). This could be the effect
blade. Both blades will see the same geometric angle of of the blade’s interaction with the large vortices that
attack and so the NACA0012 generates greater lift. were generated by other blades in the previous quarter

Figure 6. Symmetric aerofoil performance plots: (a) power coefficient curves; (b) moment coefficient plot for one full rotation of a
single blade,  ¼ 3.5; (c) pressure coefficient plots at  ¼ 90 ,  ¼ 3.5; and (d) moment coefficient plots at  ¼ 2.5.
Danao et al. 875

of rotation (Figure 7(a) to (d)). At  ¼ 210 , the torque is produced until the end of its rotation at
NACA0012 experiences separated flow, thereby losing  ¼ 360 .
lift. Drag is significantly higher for this part of the A slightly different torque ripple is observed with the
NACA0012’s rotation as well. Going past  ¼ 270 , thicker NACA0022. Again, the vortices released from
the flow over the NACA0012 reattaches and positive the second quarter of rotation of other two blades affect

Figure 7. Contours of vorticity at different blade positions for  ¼ 2.5.


Numbers 1–3 correspond to the three blades of the VAWT.
876 Proc IMechE Part A: J Power and Energy 226(7)

Figure 8. Aerodynamic performance plots: (a) lift coefficient loop for NACA0012 in one full rotation; (b) drag coefficient loop for
NACA0012 in one full rotation; (c) power coefficient curves for cambered blade profiles; (d) moment coefficient plots of cambered
blades,  ¼ 2.5; (e) lift coefficient loop for the LS0421 and inverted LS0421 profile,  ¼ 2.5; and (f) drag coefficient loop for the LS0421
and inverted LS0421 profile,  ¼ 2.5.
NACA: National Advisory Committee on Aeronautics; LS, low-speed aerofoil series.

the performance of the current blade passing the third passing blades in the third quarter. Large vortices are
quarter of rotation of the VAWT. But the difference not seen for  ¼ 3.5 throughout the full rotation of the
from the thinner NACA0012 is that the vortices are VAWT but some flow separation is detected near the
smaller and are dissipated faster. It can be seen from trailing edge for high angles of attack. Data are not
Figure 7(e) to (h) that the NACA0022 blade interacts shown for reasons of brevity. With the flow almost
with fewer vortices. Also, the flow is attached to the always attached to the blade, this is seen as one of
blade at  ¼ 210 , providing the blade an opportunity the major factors to the predominantly positive
to produce lift. There is also an observed positive torque that the blade generates.
torque generated in the last quarter of rotation. At  ¼ 4.0, it is observed that the flow is fully
At higher  the trend is clear, as the NACA0012 attached to the suction surface of the blade for more
stands out with higher performance. At  ¼ 3.0, than three quarters of the full rotation. This was also
vortex structures are observed to be generated much observed by Consul et al.9 in their 2D simulations of a
later in the second quarter of rotation and are also NACA0015. Although there is an increase in , lift
dissipated faster, hence less of them interact with the (Figure 8(a)) and power coefficient (Figure 6a) at
Danao et al. 877

 ¼ 4.0 are still lower compared to  ¼ 3.5. At high , a clear difference in the torque generated by each blade
the limiting factor to the lift generated is the low angle at just stalled conditions. The inverted LS0421 gener-
of attack seen by the blade. The drag is also seen to be ates the greatest torque while the NACA5522 has the
lower at higher  (Figure 8(b)). The torque generated at least. While the torque is high for the inverted LS0421
 ¼ 4.0 has a slightly lower peak in the upwind and in the upwind, the downwind performance is poor with
downwind compared to  ¼ 3.5. low torque values compared to most of the other blade
profiles.
This is also true for the inverted NACA5522 where
Camber effects the torque is largely positive in the upwind but mostly
The motivation for the present camber study is to deter- negative in the downwind. This is probably due to the
mine if there are advantages in the use of a cambered fact that in the upwind, the direction of the camber
aerofoil for the VAWT given that the path traversed by produces a positive angle of attack, i.e. the suction
the VAWT blades is circular, effectively giving a virtual side of the blade is the surface outside the curve of
inverted camber to the aerofoil.15 Previous studies16–20 the camber. However, in the downwind side, this rela-
have suggested that an asymmetric aerofoil improves tionship is reversed with the suction side of the blade in
the performance of the VAWT inferred from static the inside of the camber curve. As such, flow is much
aerofoil force data and mathematical models, none of more separated as the blade passes the downwind side.
these though have discussed the physics of flow in a The lift coefficient on the inverted LS0421 at  ¼ 330
detailed manner that only visualisation can provide. (Figure 8(e),  ¼ 8.7 ) is 0.209 while the drag
This is where CFD’s advantage comes in. Two coefficient is 0.007 (Figure 8(f)). Compare this to
cambered blade profiles were tested and compared to  ¼ 90 ( ¼ 21.8 ) where the lift coefficient is 0.714
the symmetric aerofoil NACA0022. A cambered and the drag coefficient is equal to 0.283; so, it is
NACA5522 aerofoil was first selected with 5% concluded that very little lift (and effectively low
camber at 50% chord position and thickness 22%c. torque) is produced in the downwind. In fact, the
Both positive camber (in the direction of, but not neces- moment coefficient predicted at  ¼ 330 is negative at
sarily on, the path curve) and inverted camber were 0.010 (Figure 8(d)). Contrary to Kirke’s19 conclusion,
tested. The next aerofoil chosen was the LS0421 with a cambered aerofoil whose concave side is facing out-
2.3% camber and a thickness of 21%c. ward (inverted camber) offers poorer performance than
Of the aerofoils tested, the positive camber LS0421 the one that has a positive camber.
performs best overall with a maximum power coeffi- It is now appropriate to examine and compare the
cient of 0.40 at  ¼ 3.5. The NACA0022 and the factors affecting force predictions over the aerofoil
inverted LS0421 have very similar power coefficient surface for the LS0421 and inverted LS0421. Shown
predictions from  ¼ 3.0–4.0. Both the NACA5522 in Figure 9 are pressure coefficient and skin friction
and the inverted NACA5522 perform poorly compared coefficient plots for the two aerofoils at six different
to the other profiles at these higher . It is worthy to azimuth positions. Initially focusing on the inverted
mention that the power coefficient profile of the LS0421 profile in the upwind region, a clear distinc-
NACA5522 is the flattest. This is a favourable feature tion can be made between the three azimuth positions
in unsteady wind conditions because a change in  does in so far as the size of the enclosed area of the pres-
not significantly change the power coefficient. However sure coefficient curve is concerned. Between  ¼ 60
with the same flow conditions, the power derived by and  ¼ 90 , there is not much distinguishable differ-
this blade is considerably lower than the other aerofoil ence in the size of the enclosed area. However, the
profiles. Both inverted camber profiles (NACA5522-inv skin friction coefficient at  ¼ 60 is generally higher,
and LS0421-inv) perform poorly at low . thereby increasing the drag and subsequently decreas-
Positive cambered aerofoils present an advantage at ing the torque produced as evidenced by the lower
low , as seen in Figure 8(c). A camber along the cir- moment coefficient.
cular path of the VAWT blade as with the NACA5522 Shifting attention to the LS0421 in the downwind,
and the LS0421 gives higher power coefficient values the pressure coefficient plots show that it is at  ¼ 330
versus the straight NACA0022 at  ¼ 2.0. Giving the where the enclosed area is greatest amongst the three.
blade a slight camber improves the overall performance However, the skin friction at the same azimuth position
of the VAWT which is also seen by Beri and Yao16 and is also the greatest between the three. These three azi-
Baker.17 This does not completely contradict the find- muth positions have roughly the same moment coeffi-
ings of Healy18 but suggests that a symmetric aerofoil is cient values regardless of their differences in pressure
not the best profile when it comes to VAWT perform- and skin friction values. The inverted camber is ideal in
ance. At  ¼ 2.5, a close inspection at the moment the upwind while the positive camber is preferred in the
coefficient plot just before  ¼ 90 (Figure 8(d)) shows downwind.
878 Proc IMechE Part A: J Power and Energy 226(7)

Figure 9. Moment coefficient plots for one blade of the LS0421 and inverted LS0421 profiles, one full rotation at  ¼ 3.5, with
pressure coefficient and skin friction coefficient plots at different stages of the rotation.
LS, low-speed aerofoil series.

Overall, the positive camber gives better perform- generate high values of torque over the entire rotation,
ance than the inverted camber profile. The advantage giving it a higher mean torque and consequently higher
of the inverted camber in the upwind is drastically CP. Over a wide range of , the positive camber is still
reduced by its very low performance in the downwind. the better performing profile with considerably
Whereas the inverted camber’s performance is accept- improved performance over the inverted camber in
able only in the upwind, the positive camber is able to the low  range.
Danao et al. 879

Figure 10. Moment coefficient plots for one blade of the NACA0022, NACA5522 and LS0421 profiles, one full rotation at  ¼ 3.0,
with pressure coefficient plots and pressure contour plots superimposed with streamlines at different stages of the rotation.
NACA: National Advisory Committee on Aeronautics; LS, low-speed aerofoil series.
Notes: Pressure coefficient plots do not have the same scales in the ordinate axis.

To complete the analysis for the camber study, there At first glance, there is not much difference in the torque
is a need to compare the performance of a cambered ripple between the three profiles except for the region
aerofoil to a symmetric profile. Shown in Figure 10 approaching  ¼ 150 where the torque values for the
are the moment coefficient plots for three aerofoil sec- NACA5522 deviate from the other two and plunge
tions, NACA0022, NACA5522 and LS0421, at  ¼ 3.0. into the negative zone. The differences in power
880 Proc IMechE Part A: J Power and Energy 226(7)

coefficients between the three do not exceed 6%. In fact, thickness and camber variation. Results show that for
the difference between the NACA0022 and the LS0421 is the two symmetric aerofoils tested, the thinner blade is
only 1%. But it is at this tip speed ratio where the great- more preferred with maximum CP up to 0.50 for
est gap in power coefficients is observed, hence the NACA0012 at  ¼ 3.5. Over the entire range of 
choice for this part of the analysis. tested, the NACA0012 aerofoil performs better than
For most of the rotation, the flow is partially to fully the NACA0022. Slightly cambered aerofoils such as
attached to the blades. At maximum torque close to LS0421 profile can improve the overall performance
 ¼ 90 , it can be seen from the streamline plot that the of the VAWT whereas a camber of 5% as the case
flow is almost fully attached to the NACA5522 and par- for NACA5522 is unfavourable. The cambered
tially attached to the NACA0022 and LS0421. However, NACA5522 is, in general, inferior to the symmetric
the NACA0022 generates the largest torque of the three NACA0022, especially in the high  region. Of the
at this particular stage. Visual inspection of the flow field thicker blades tested, the LS0421 performs the best
cannot directly tell us which blade is performing better. with maximum CP of 0.40 at  ¼ 3.5. A camber along
This is a true observation to other stages in the rotation the blade path causes the blades to generate high values
as well. At  ¼ 240 , the flow is almost fully attached to of torque in both the upwind and the downwind
the NACA0022 while less attached to the other two. In regions. On the other hand, inverted cambered profiles
spite of this, it is the LS0421 that generates the greatest produce power only in the upwind side. Between the
torque at this azimuth position. candidate profiles investigated, the NACA0012 gave
It is at  ¼ 150 where a definite conclusion can be the best overall aerodynamic performance for power
made from the given plots. As mentioned above, the extraction. CFD flow visualisations complemented by
torque values for the NACA5522 drop into the negative a non-traditional discussion of the physics of flow offer
region while the other two profiles’ hover close to zero. new insight into the way VAWT performance analysis
The pressure coefficient plot for this azimuth shows is carried out.
that the surface of the NACA5522 switch from being
the suction side to the pressure side and moving back Funding
from the leading edge to the trailing edge. This switch is This work was funded by the Engineering Research and
indication of loss of lift as the aerofoil experiences deep Development for Technology Program of the Department
stall. It is unmistakable in the streamline plot that there of Science and Technology through the University of the
is a large-scale vortex that is shed by the NACA5522 Philippines – College of Engineering.
while none can be seen from the NACA0022 or the
LS0421.
Throughout the whole rotation, only the References
NACA5522 produced large vortices as the blades pass 1. UK energy in brief. National statistics publication. URN
the second quarter of revolution. The vortices shed are 10D/220, July 2010, http://www.deec.gov.uk (accessed 26
not associated to dynamic stall and as such do not August 2010).
positively contribute to the lift generated. They are 2. American Wind Energy Association. Vertical axis wind
seen to develop from the trailing edge as a phenomenon turbine: the history of the DOE program. US
Department of Energy, Sandia National Laboratories,
that can accompany stalled and separated flow condi-
American Wind Energy Association, 1986.
tions. Overall, the LS0421 is the best performing aero- 3. Dodd HM. Performance predictions for an intermediate-
foil of the cambered profiles tested over a wide range of sized VAWT based on performance of the 34-m VAWT
tip speed ratios. It is able to generate reasonable values test bed. In: Proceedings of the 9th ASME wind energy
of torque in both the upwind and the downwind, symposium (ed DE Berg), New Orleans, LA, 14–17
thereby increasing its mean power coefficient. January 1990, vol. 9.
4. Dodd HH, Ashwill TD, Berg DE, et al. Test results and
status of the DOE/Sandia 34-m VAWT test bed. In:
Conclusions Proceedings of the Canadian wind energy association con-
A pitching aerofoil study was performed to determine ference, Albuquerque, NM, 14 September 1989.
the validity of a set of turbulence models for VAWT 5. Berg DE, Klimas PC and Stephenson WA. Aerodynamic
design and initial performance measurements for the
simulations. Very good agreement in both force and
Sandia 34-m vertical axis wind turbine. In: Proceedings
flow predictions were seen with the use of the SST k–! of the 9th ASME wind energy symposium, New Orleans,
model. It is shown that the SST k–! model is the most LA, 14–17 January 1990, vol. 9.
appropriate turbulence model for use in unsteady 6. Price TJ. UK large-scale wind power programme from
aerofoil motion with dynamic stall phenomenon. 1970 to 1990: the Carmarthen Bay experiments and the
2D CFD simulations were carried out to investigate Musgrove vertical-axis turbines. Wind Eng 2006; 30(3):
the performance effects of VAWT blade aerofoil 225–242.
Danao et al. 881

7. Hamada K, Smith T, Howell R, et al. Unsteady flow


simulation and dynamic stall around vertical axis wind Appendix
turbine blades. In: 27th ASME wind energy symposium,
Reno, NV, 7–10 January 2008, AIAA-2008–1319. Notation
8. Howell R, Qin N, Edwards J, et al. Wind tunnel and
c aerofoil chord
numerical study of a small vertical axis wind turbine.
Renew Energy 2010; 35: 412–422.
Cd drag coefficient
9. Consul CA, Willden RHJ, Ferrer E, et al., Influence of Cf skin friction coefficient
solidity on the performance of a cross-flow turbine. In: Cl lift coefficient
Proceedings of the 8th European wave and tidal energy Cm moment coefficient
conference, Uppsala, Sweden, 7–10 September 2009. Cp pressure coefficient
10. Ferreira CS, Van Bussel G and Van Kuik G. 2D CFD CP power coefficient
simulation of dynamic stall on a vertical axis wind tur- d VAWT rotor diameter
bine: verification and validation with PIV measurements. D drag force
In: 45th AIAA aerospace sciences meeting and exhibit, k turbulence kinetic energy
Reno, NV, 8–11 January 2007, AIAA-2007–1367. l turbulence length scale
11. Lee T and Gerontakos P. Investigation of flow over an
L lift force
oscillating aerofoil. J Fluid Mech 2004; 512: 313–341.
M moment
12. Fluent. Fluent 6.3 user’s guide, Chapter 12, 2006.
13. Healy JV. The influence of blade thickness on the output r VAWT rotor radius
of vertical axis wind turbines. Wind Eng 1978; 2(1): 1–9. Rec chord-based Reynolds number
14. Jacobs E and Sherman A. Airfoil section characteristics as Tu turbulence intensity
affected by variations of the Reynolds number. NACA U1 free-stream wind velocity
Report no. 586, 1937. V blade velocity
15. Migliore PG, Wolfe WP and Fanucci JB. Flow curvature W blade relative velocity
effects on Darrieus turbine blade aerodynamics. J Energy yþ dimensionless wall distance
1980; 4: 49–55.
16. Beri H and Yao Y. Effect of camber airfoil on self start-  instantaneous angle of attack
ing of vertical axis wind turbine. J Environ Sci Technol
A pitching magnitude
2011; 4: 302–312.
O mean angle of attack
17. Baker JR. Features to aid or enable self-starting of fixed
pitch low solidity vertical axis wind turbines. J Wind Eng " turbulent dissipation rate
Ind Aerodyn 1983; 15: 369–380.  azimuth angle
18. Healy JV. The influence of blade camber on the output of  reduced frequency
vertical axis wind turbines. Wind Eng 1978; 2(3): 146–155.  tip speed ratio
19. Kirke BK. Evaluation of self-starting vertical axis wind  air density
turbines for stand-alone applications. PhD Thesis, ! specific dissipation rate
Griffith University, Australia, 1998.
rotor angular velocity
20. Kadlec EG. Characteristics of future vertical axis wind
turbines. Sandia National Laboratory, Albuquerque,
NM, November, 1982.

Potrebbero piacerti anche