Aerodynamic Analysis of a Vertical Axis Wind Turbine

in a Diffuser
Ben Geurts Carlos Simão Ferreira Gerard van Bussel
[email protected] [email protected] [email protected]

DUWIND
Delft University of Technology
Kluyverweg 1
2629HS Delft
The Netherlands

Abstract 1 The Operation of Diffuser-

augmented Wind Turbines
(DAWTs)

The amount of energy that can be extracted by

a HAWT is expressed as the non-dimensional
power coefficient CP . It is the ratio between the ki-
Wind energy in the urban environment faces com- netic power which is contained in the free-stream
plex and often unfavorable wind conditions. High wind and the amount of energy that is eventually
turbulence, lower average wind velocities and converted by the wind turbine. HAWTs can be
rapid changes in the wind direction are common modeled in a very simplistic way as actuator discs
phenomena in the complex built environments. in a quasi 1-dimensional flow. This actuator disc
A possible way to improve the cost-efficiency of model shows that the maximum obtainable power
urban wind turbines is the application of flow- coefficient is 0.59. This theoretical maximum is
enhancing structures on or near the turbines. For obtained when the axial induction factor a is equal
horizontal axis wind turbines (HAWTs), applying a to 1/3. In this situation the flow is decelerated
diffuser has shown to have a beneficial impact on to 2/3 and 1/3 of its original free-stream velocity
the power production, but it is still under develop- at the rotor-disc and in the far wake respectively.
ment. For a vertical axis wind turbine (VAWT) it is As explained by Van Bussel [3],[4], this same ra-
expected that flow augmentation will also strongly tio also applies to a diffuser-augmented wind tur-
increase the performance of the turbine, but very bine (DAWT) as long as the diffuser does not in-
little research has been done in this field. The duce a (negative) back-pressure. In this analogy
purpose of this research is to investigate the ef- with unaugmented turbines, the flow deceleration
fects of a diffuser on the airflow through a VAWT. of 1/3 is now assumed to occur at the exit of the
In order to investigate these effects, the turbine diffuser and can be expressed as in equation 1.
(with and without diffuser) is simulated using a 2- With this same velocity reduction ratio, the power
D unsteady free-wake potential-flow panel model. increase due to the diffuser is purely a result of the
The local flow field, local angles of attack, shed increased mass flow trough the rotor-disc. The
vorticity, the shape and strength of the wake, and area of the rotor-disc remains the same, but as a
the rotor torque are investigated for both the case result of the diffuser, the flow has a higher velocity,
with and without the diffuser. The diffuser used and as a consequence, the 1/3 velocity reduction
in this research consists of two mirrored airfoil is applied to a larger air-mass flowing through the
cross-sections. The size of the duct-opening in rotor. The increase of velocity is directly related to
which the turbine operates is varied. This work the ratio β of the diffuser exit area A3 with respect
shows that unlike for a 1-D actuator disc analysis, to the diffuser nozzle area A1 as given in equa-
the area ratio β of the diffuser exit with respect tion 2. Combining equation 1 and 2, the velocity
to the diffuser nozzle area is not the only driving V1 in the nozzle of the diffuser can be expressed
factor in the augmentation of the rotor torque of as a function of the free-stream velocity V0 as in
the VAWT. More important are the effect of the di- equation 3.
rectional change of the rotor inflow and the faster
downstream transport of the shed vorticity. V3 = (1 − a)V0 (1)
 A3
V1 = βV3 = V3 (2) words: it provides a DAWT-equivalent of the well-
A1 known Betz-limit.
V1 = β(1 − a)V0 (3) In order to evaluate the aerodynamic behavior of
a VAWT and the effect of diffuser-augmentation,
When the increased power is expressed as a a 2-D simulation code is used. The code is an
power coefficient with respect to the rotor exit unsteady free-wake potential-flow panel model as
area, the maximum obtainable power coefficient described and used in [2]. It is this M ATLAB-based
of 0.59 still holds. However, when related to swept program which is used to obtain the results dis-
area of the rotor in the nozzle of the diffuser, the played and discussed in the following sections.
maximum CP can surpass this value. The blades and diffuser airfoils are modeled as
doublet panels, while the wake is represented by
free-moving singular vortex points.

3 Basic Vertical Axis Wind Tur-

bine Operation
In steady operation, the aerodynamic loading on a
VAWT is a periodic effect with the frequency of the
rotation. This is much different from the steady op-
eration of a HAWT, where the aerodynamic load-
Figure 1: Pressure and velocity relations in an
ing on the blades remains steady throughout the
empty diffuser without back-pressure
rotation. There are different ways to present the
rotational variation of the loading on a VAWT since
practically all output-variables experience related
Figure 1 shows the pressure and velocity for an
fluctuations. Figure 3 shows the instantaneous
empty diffuser without back-pressure. It is possi-
angle of attack experienced by a single blade over
ble, however, that a back-pressure exists because
one rotor revolution. As a result of the contin-
of the Kutta condition, when the flow is forced to
uously changing angle of attack, the forces on
deflect in radial direction. As a result, the veloc-
the blade change accordingly. In figure 4, the ra-
ity at the exit of the empty diffuser V3 can differ
dial and tangential components of the force co-
from the undisturbed flow velocity far in front of
efficients on one single blade are given. Time-
the diffuser V0 as given in equation 4. The ve-
dependent loading on the blade section causes
locity increase results in a further augmentation
the shedding of vorticity. As can be seen in figure
of the power produced by a turbine placed in the
5, the shed vorticity obtains maximum strengths
nozzle of the diffuser.
around azimuth positions of 35◦ (negative vortic-
V3 = γV0 (4) ity) and 170◦ (positive vorticity). The shed vorticity
is transported into the wake, as visualized in fig-
The maximum achievable (ideal) power coeffi- ure 6. In figure 7, a flow field of streamlines is
cient CP can be expressed as in equation 5 or shown resulting from the time-averaged local ve-
equation 6 when it is based on the rotor disc area locities over one rotor revolution.
or the diffuser exit area respectively.

CProtor = βγ4a(1 − a)2 (5)

4 Diffuser-Augmented VAWT
CPexit = γ4a(1 − a)2 (6) In this section, the results from the previous sec-
tion are compared with the simulation results of
the same VAWT operating in the nozzle of a dif-
2 Simulation Method fuser. For this general comparison, the area of the
diffuser nozzle is equal to 1.6 times the diameter
As explained by Simão Ferreira [2], the 1-D ac- of the rotor.
tuator disc model is not very well-suited for mod- The actual operation of the VAWT inside a diffuser
eling a VAWT. However, it only assumes a pres- varies little from the free-stream operation. The
sure drop over the disc representing the energy presence of the diffuser provides a flow acceler-
extraction by the turbine and therefore it gives a ation through the rotor. With a fixed tip speed,
rough indication of the theoretic maximum power the local angles of attack experienced by a blade
increase as a result of a diffuser. Or in other during one revolution increase, as can clearly
 Figure 5: Normalized strength of the vorticity shed
∂Γwake
by a single blade per timestep ( ∂θcπU )
Figure 2: Azimuth position definition for a clock- ∞

wise rotating VAWT

Figure 6: Phase-averaged vorticity distribution

Γ c
c2 U∞ with the rotor at the zero azimuth position
Figure 3: Phase-averaged angle of attack vs az-
imuthal position

Figure 7: Streamlines visualizing the time-

averaged local velocity field
Figure 4: Instantaneous radial and tangential
force coefficient of one single blade ( 1 ρλ2FU 2 c )
2 ∞
these forces are given in figures 9 and 10. Note
that the tangential force coefficient is positive in
be seen in figure 8. These higher absolute val- the direction of the rotation.
ues for the angle of attack induce a higher blade When looking at this tangential force coefficient,
loading, both in radial and in tangential direc- it is easy to see that there is a significant increase
tion. The dimensionless coefficients representing of rotor torque as a result of the diffuser. The time-
averaged power coefficient Cp for the diffuser-
unaugmented turbine is 1.72 times as large as that
of the un-augmented turbine.Which means an in-
crease of about 72 percent. In order to make the
comparison with the power increase as predicted
by the 1-D actuator disc theory, the diffuser area
ratio has to be evaluated. For the diffuser used
here, the area ratio of the diffuser with respect to
the nozzle is 1.76. This implies a 1-D prediction of
power increase of 76 percent assuming no nega-
tive back-pressure.
This means the predicted power increase pro-
vided by the diffuser agrees almost perfectly with
the obtained rise in power from the simulations. In Figure 10: Tangential force coefficient, contribu-
order to validate the comparison made here, the tion of one single blade ( 1 FT 2 )
2
2 ρλ U∞ c
same VAWT is next fitted with different diffusers,
having different area ratios.
5 Augmenting Effects of the
Diffuser
The VAWT is simulated in combination with dif-
ferent diffuser areas. For every setup, the same
cross-sectional shape is used for the diffuser.
With the increasing nozzle area, the exit area ra-
tio decreases, and since the diffuser is located
further from the VAWT, its influence is expected
to decrease. We look at the cases were the dif-
fuser nozzle is 1.2, 1.4, 1.6, 1.8 and 2 times the
diameter of the turbine. Four of these setups are
shown in figure 11. The tangential force coeffi-
cient for all the above-mentioned cases is shown
in figure 12. The average power coefficient Cp
Figure 8: Angle of attack comparison between can be expressed relative to the power coefficient
free and diffuser-augmented VAWT of the unaugmented VAWT, as is given in table
1 along with the ratio β of the diffuser exit area
with respect to the nozzle area. As can clearly be
concluded from table 1, there is hardly any torque
increase as a result of the 1.2R-diffuser with the
highest area ratio β. This fact suggests that the
area ratio is not the driving factor in the diffuser-
augmented operation of the VAWT, and the good
agreement of augmentation factors in the previous
section can be considered to be a lucky shot.

Table 1: Comparison of diffuser exit area ratio β

C
and relative power coefficient Cpp with Cp0 the
0
power coefficient of the unaugmented turbine
Anozzle β relative Cp
No Diffuser - 1
1.2R 2 1.06
Figure 9: Radial force coefficient, contribution of 1.4R 1.87 1.23
FR
one single blade ( 1 ρλ2 U 2 c ) 1.6R 1.76 1.72
∞
2 1.8R 1.67 1.66
2R 1.61 1.80
 With increasing diffuser nozzle area, the
decrease of the peak in the tangential force
coefficient plot at 100◦ transforms into an increase
with respect to the unaugmented case providing a
steady rise of the torque at all azimuth positions.
This trend also shows when looking at the local
angle of attack in figure 13.

In order to understand what is happening, we

should take a look at the local velocity of the flow
through the rotor. Figures 14 and 15 show the
local time-averaged induction a = 1 − UU∞ x
in the
direction of the free-stream flow. The large local
angle of attack in the case of the 1.6R diffuser
is caused by a higher incoming flow velocity at
the rotor. This is a striking observation, since the
velocity increase inside the empty 1.2R diffuser
is expected to be higher than that of the 1.6R
diffuser. This clearly means that the prediction
of maximum achievable power increase by the
1-D actuator disc theory does not hold for a
diffuser-augmented VAWT. The reason for this
can be found in the gap between the inside of the
diffuser-nozzle and the turbine. For the diffuser
Figure 11: Visualization of the VAWT-diffuser with the 1.2R nozzle section, there is practically
combination for four of the simulated cases no gap. The shape of the diffuser airfoil causes
the peak in the local angle of attack to occur
earlier in the rotation, and to remain until a later
azimuth position than the unaugmented turbine.
This is due to the local flow direction rather than
due to the local flow velocity. This directional
difference near the azimuth positions of 35◦ and
160◦ is visible when comparing the unaugmented
flow field in figure 7 with the augmented flow
field in figure 16. The change in shape of the
load-peak in the up-wind part of the rotation also
has consequences for the shedding of the vor-
ticity. Figure 17 shows that the peak of the shed
vorticity of the 1.2R diffuser is higher and occurs
in a shorter time-span then for the unaugmented
turbine. The stronger and more concentrated vor-
tices that are shed result in a stronger induction.
Around the 90◦ azimuth position, the increased
Figure 12: Torque coefficient contribution of induction reduces the local flow velocity, and thus
single blade for different diffuser nozzle areas decreases the local angle of attack experienced
( 1 ρλF2 UT 2 c ) by the blades. This explains the power loss which
2 ∞

can be seen as the lower tangential force peak in

From the torque-contribution of a single blade figure 12 with respect to the unaugmented case.
in figure 12, it is possible to see that there is a
drop in the peak of the tangential force around But then why does this not happen in the same
the azimuth position of 100◦ for the 1.2R diffuser way when a larger diffuser-nozzle is applied? The
case. Also, there is a gain around 35◦ and 160◦ . answer to this question might be found in the rep-
These phases of power gain cancel out with the resentations of the vorticity distribution. Figures
drop of the power peak, resulting in almost the 18 and 19 show the distribution of vorticity for the
same average power coefficient Cp over a full 1.2R and the 1.6R diffuser respectively. For the
rotation as the unaugmented turbine (see table 1). 1.6R diffuser, due to the faster-moving flow just
outside the rotor, the concentrated vortices which
are shed with every blade passing are transported
faster downwind. The amount of vorticity shed
per blade passage in this case is also larger as
could be seen in figure 17. This larger vortex
strength has a stronger induction at the rotor as
a consequence. However, the effect of a faster
vortex-movement in the case of the 1.6R has a
stronger influence. When the shed vortices are
further away from the rotor, they induce less veloc-
ity decrease at the rotor, allowing for higher posi-
tive angles of attack at the upwind blade positions
and negative angles at the downwind blade posi-
tions, both resulting in a larger torque on the rotor
over the whole rotation.
Figure 15: Time-averaged local velocity induction
a = 1− UU∞
x
for the VAWT operating inside the 1.6R
diffuser

Figure 13: Local angle of attack for different dif- Figure 16: Streamlines visualizing the time-
fuser nozzle areas averaged local velocity field

Figure 17: Normalized strength of the shed vortic-

∂Γwake
ity of a single blade ( ∂θcπU ∞
)

Figure 14: Time-averaged local velocity induction

6 Conclusions and Further In-
a = 1− UU∞
x
for the VAWT operating inside the 1.2R vestigations
diffuser
For the evaluation of diffuser-augmented horizon-
tal axis wind turbines (DAWT’s), a 1-D actuator
 of attack, providing more torque. Thus, rather
than the area ratio of the diffuser itself (and thus
the velocity increase produced by the empty
diffuser) it is the size of the gap between the
rotor and the diffuser that effects the power
augmentation of the diffuser-augmented VAWT.
In that context, the work of Aquiló et al ([1]) should
be revisited. Aquiló et al use CFD-calculations
to predict the (1-D) speed-up of the flow through
a building-augmented wind concentrator. For the
application of HAWTs, this estimation might proof
valid, but when this approach is used to estimate
the expected power augmentation of a VAWT
Figure 18: Phase-averaged vorticity distribution placed inside this wind concentrator system, the
Γ c
c2 U∞ with the rotor at the zero azimuth position result may become invalid.
inside the 1.2R diffuser
Even though the presented aerodynamic
investigation gives a clear explanation for the
occurring phenomena, it is considered that
other aerodynamic effects might play a role in
reducing or enhancing the explained effects. To
this extend, more effort is needed in order to
investigate other possible effects influencing the
performance of the augmented VAWT. Since the
area ratio of the diffuser is not the driving factor in
the augmentation, a whole different diffuser setup
should be considered. On the other hand, the
above conclusions should be checked and proven
using other simulation methods and experimental
research.
Figure 19: Phase-averaged vorticity distribution
Γ c
c2 U∞ with the rotor at the zero azimuth position
inside the 1.6R diffuser
References
disc theory can be used linking the theoretical [1] A. Aguilo, D. Taylor, A. Quinn, and R. Wilt-
maximum power increase of the DAWT to the shire. Computational fluid dynamic mod-
velocity increase produced by the empty diffuser. elling of windspeed enhancement through a
The velocity increase, and thus also the power building-augmented wind concentration sys-
increase, can be directly expressed as the ratio tem. EWEC Conference 2004 - Poster Pre-
of the exit area of the diffuser with respect to the sentation, 2004.
nozzle area of the diffuser. Although not perfectly [2] Carlos Simão Ferreira. The near wake of
suitable for application to a vertical axis wind the VAWT. PhD thesis, Faculty of Aerospace
turbine (VAWT), this 1-D approach can serve Engineering, Delft University of Technology,
as a reference for the expected power increase. 2009.
From initial simulations with a nozzle area of
1.6 times the rotor diameter, it can be seen that [3] Gerard J. W. van Bussel. An assessment of
the power increase of the VAWT nicely fits the the performance of diffuser augmented wind
increase predicted by the 1-D actuator theory. As turbines (dawt’s). 3rd ASME/JSME Fluid En-
is shown by varying the nozzle area (1.2, 1.4, 1.6, gineering Conference FEDSM99-7830, San
1.8 and 2 times the rotor diameter), the amount Fransisco, USA, 1999.
of power increase is dependent on the size of
the gap between the rotor and the diffuser. The [4] Gerard J. W. van Bussel. The science of mak-
high-velocity flow outside the rotor will transport ing more torque from wind: Diffuser experi-
the shed vorticity away from the rotor, reducing ments and theory revisited. The Science of
the local induction at the blade positions. The Making Torque from Wind, Journal of Physics:
lower local induction allows for larger local angles Conference Series 75, 2007.