You are on page 1of 20

ARTICLE IN PRESS

Journal of Wind Engineering


and Industrial Aerodynamics 93 (2005) 951970
www.elsevier.com/locate/jweia

Experimental and numerical investigations


of ow elds behind a small wind turbine
with a anged diffuser
K. Abea,, M. Nishidab, A. Sakuraia, Y. Ohyac, H. Kiharaa,
E. Wadad, K. Satod
a

Department of Aeronautics and Astronautics, Kyushu University, Hakozaki, Higashi-ku, Fukuoka 812-8581, Japan
b
Department of Aerospace Systems Engineering, Sojo University, Ikeda, Kumamoto 860-0082, Japan
c
Research Institute for Applied Mechanics, Kyushu University, Kasuga-Kouen, Kasuga 816-8580, Japan
d
Graduate student, Department of Aeronautics and Astronautics, Kyushu University, Hakozaki,
Higashi-ku, Fukuoka 812-8581, Japan
Received 5 October 2004; received in revised form 8 September 2005; accepted 19 September 2005
Available online 27 October 2005

Abstract
Experimental and numerical investigations were carried out for ow elds of a small wind turbine
with a anged diffuser. The present wind-turbine system gave a power coefcient higher than the
Betz limit ( 16=27) owing to the effect of the anged diffuser. To elucidate the ow mechanism,
mean velocity proles behind a wind turbine were measured using a hot-wire technique. By
processing the obtained data, characteristic values of the ow elds were estimated and compared
with those for a bare wind turbine. In addition, computations corresponding to the experimental
conditions were made to assess the predictive performance of the simulation model presently used
and also to investigate the ow eld in more detail. The present experimental and numerical results
gave useful information about the ow mechanism behind a wind turbine with a anged diffuser. In
particular, a considerable difference was seen in the destruction process of the tip vortex between the
bare wind turbine and the wind turbine with a anged diffuser.
r 2005 Elsevier Ltd. All rights reserved.
Keywords: Wind turbine; Flanged diffuser; Tip vortex; Separation; CFD; Turbulence; Non-linear Eddy-viscosity
model; Disk-loading method

Corresponding author. Tel.: +81 926 423723; fax: +81 926 423752.

E-mail address: abe@aero.kyushu-u.ac.jp (K. Abe).


0167-6105/$ - see front matter r 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jweia.2005.09.003

ARTICLE IN PRESS
952

K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

1. Introduction
The power in wind is well known to be proportional to the cubic power of the wind
velocity approaching the wind turbine. This means that even a small amount of
acceleration gives a large increase in the energy output. Therefore, many research groups
have tried to nd a way to accelerate the approaching wind effectively [17].
Recently, Ohya et al. [8,9] have developed an effective windacceleration system.
Although it adopts a diffuser-shaped structure surrounding a wind turbine like the others
previously proposed, the feature that distinguishes it from the others is a large ange
attached at the exit of diffuser shroud. Fig. 1 illustrates an overview of the present
windacceleration system. A ange generates a large separation behind it, where a very
low-pressure region appears to draw more wind compared to a diffuser with no ange.
Owing to this effect, the ow coming into the diffuser can be effectively concentrated and
accelerated. In this system, the maximum velocity is obtained near the inlet of diffuser and
thus a wind turbine is located there as shown in Fig. 1. Although this windturbine system

(a)

(b)
Fig. 1. Schematic view of a diffuser-shrouded wind turbine: (a) overview of system; (b) ow mechanism around a
anged diffuser.

ARTICLE IN PRESS
K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

953

has already shown some promising results [8,9], more detailed investigations of ow elds
inside the diffuser are still needed to achieve the best performance.
Having considered the above background, in this study experimental and numerical
investigations are carried out for ow elds of a small wind turbine with a anged diffuser
(i.e. a diffuser-shrouded wind turbine). In the experiments, mean velocity proles behind
the wind turbine are measured using a hot-wire technique. Furthermore, computations
corresponding to the experimental conditions are made to assess the predictive
performance of the simulation model presently used and also to investigate the ow eld
in more detail. By processing the obtained data, characteristic values of the ow elds are
estimated and compared with those for a bare wind turbine.
2. Experimental apparatus
All measurements were performed in a large wind tunnel at the Department of
Aeronautics and Astronautics, Kyushu University. It has a measurement section of 2.5 m
(width) 1:5 m (height), with a maximum wind velocity of 20 m/s. In this study, a wind
turbine is located 500 mm downstream of the wind-tunnel exit. The wind turbine was a
three-bladed rotor with a diameter of 388 mm. The blade was designed for the diffusershrouded wind turbine currently tested, with the aid of the design theory developed by
Inoue et al. [10].
Fig. 2 gives detailed information on the diffuser-shrouded wind turbine. The diffuser
consists of a main diffuser, a ange attached at the rear of the diffuser and an inlet shroud
attached at the front. As shown in Fig. 2, the diameter of the diffuser throat was 400 mm,
where the maximum velocity was obtained and thus where the wind turbine was located.
The length of the main diffuser was 500 mm and the ange height was 200 mm. In this

Fig. 2. Detailed information of the present diffuser-shrouded wind turbine.

ARTICLE IN PRESS
954

K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

experiment, the diameter of the center hub was 80 mm and hence the hub/throat ratio was
0.2. An AC servomotor was installed in the hub to change the rotational velocity of the
wind turbine arbitrarily. Also, a torque meter was installed to measure the axial torque
generated by the wind turbine.
Fig. 3 shows the traverse system used in the present experiment. In this study, a
cylindrical coordinate system was adopted, with x, r and y being the streamwise, radial and
rotational coordinates, respectively. The origin was set at the location of the blades as
shown in Fig. 3(b). The velocity elds behind the wind turbine were measured using a hotwire technique. In this study, a hot-wire X-probe (KANOMAX) was used, each of which
consists of a tungsten wire 4 mm in diameter and 5 mm in length. Since an X-probe can
provide only two components of a velocity, we performed measurements for both xy and
xr components independently to obtain three-dimensional mean velocity elds. As seen in
Fig. 3(a), in order not to disturb the ow eld, the probe was inserted with a supporting
bar 500 mm in length. Figs. 3(b) and 4 illustrate schematic images of the measurement

Fig. 3. Traverse system: (a) overview of system; (b) detailed information.

ARTICLE IN PRESS
K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

955

Fig. 4. Schematic image of measurement: (a) bare wind turbine; (b) diffuser-shrouded wind turbine.

positions. In the experiments, to measure velocity elds just behind the wind turbine, we
rst set the hot wire at x 48 mm (just 10 mm behind the root of the blade) and r 55 mm
(15 mm away from the hub surface). From this position, the probe was traversed in the
radial direction and velocity data were obtained in the ry plane (see Fig. 3(b)). Moreover,
to obtain velocity distributions for the whole ow region, the probe was traversed in the
streamwise direction (see Fig. 4).
The free stream velocity (U 0 ) was specied as 11 and 6.8 m/s for the bare wind turbine
and the diffuser-shrouded wind turbine respectively. The latter condition (6.8 m/s for the
diffuser-shrouded wind turbine) was chosen so that the approaching wind velocity could be
similar to that in the case of the bare wind turbine. The Reynolds number based on the
diameter of the wind turbine and the free stream velocity was about 2:9  105 for the bare
wind turbine and 1:8  105 for the diffuser-shrouded wind turbine. The measurement at a
given position was started by a trigger signal detected when a selected blade passed a laser
beam. The data-sampling rate was set to between 2 and 10 kHz so as to be enough to

ARTICLE IN PRESS
K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

956

resolve in detail the rotational direction. The resolution was less than 1 (i.e. more than 360
samples/rotation) in each case. The same procedure was repeated 150300 times and the
obtained data were ensemble-averaged (i.e. phase-averaged) to calculate mean values at
each point.
3. Numerical method and computational conditions
The present ow eld is generally expressed by the continuity and the incompressible
Reynolds-averaged NavierStokes equations as follows:
qU i
0,
qxi
qU i
qP
q
rU j


qxi qxj
qxj

(1)
 


qU i qU j
rn

 rui uj F i ,
qxj
qxi

(2)

where denotes a Reynolds-averaged value. In the equations, r, P, U i , ui and n


respectively denote the density, mean static pressure, mean velocity, turbulent uctuation
and kinematic viscosity. In Eq. (2), F i is the body-force term imposed for the
representation of a load.
The present computational procedure is almost the same as that used in the previous
report by Abe and Ohya [11], with some modications in specifying the newly incorporated
load. The grid system and computational conditions are shown in Fig. 5. In this study, the
ow was assumed to be an axisymmetric steady ow, with x and r being the streamwise
and radial coordinates respectively. In Fig. 5, L, D, f and h respectively denote the diffuser
length, diffuser diameter at the throat, diffuser opening angle and height of ange used in
the present experiment, i.e., L=D 1:25, h=D 0:5 and f 12 . In Fig. 5(a), the
subscripts 0, 1, 2 and b denote values at the inlet (free stream), in front of the load
(approaching), behind the load and at the exit of the diffuser, respectively. Note that the
computational domain was determined so as not to create any serious problems for the
obtained results [11].
In this study, the load inside the diffuser was represented by the following general
expression [11]:
Ct 1
rU x jU x j; F r 0,
(3)
D2
where U x is the streamwise velocity. In Eq. (3), C t and D are the loading coefcient and its
streamwise width imposed, respectively. In this study, C t was determined by use of a diskloading method (see for example, [12,13]), as illustrated in Fig. 6. It was found that C t can
be evaluated in the two-dimensional cylindrical (xr) coordinate as follows:
Fx

ncC L cos b C D sin bf1 ro=U x 2 g


,
(4)
2pr
where n is the number of blades, o is the angular velocity of the wind turbine, c is the chord
length and b tan1 U x =ro. In Eq. (4), C D and C L are, respectively, the drag and lift
coefcients for the relative angle of attack, a b  g, where g is the angle of blade setting
(see Fig. 6). In this study, the proles of C D and C L were taken from the database [10] for
two-dimensional blade sections adopted for the wind turbine. They are shown in Fig. 7.
Ct

ARTICLE IN PRESS
K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

957

(a)

(b)
Fig. 5. Grid system and computational conditions: (a) grid system; (b) computational conditions.

On the other hand, from the relation shown in Fig. 6, the torque generated by the blades
can be estimated as follows:
Z r0
Z r0
1
rfU 2x ro2 gC L sin b  C D cos bncr dr
T
dT
rh
rh 2
Z r0
1
rU 2x C L sin b  C D cos bf1 ro=U x 2 gncr dr

rh 2
Z r0
1
r
U 2x C Z 2pr2 dr,
5
2 rh
where r0 and rh denote the radius of blade and hub, respectively. Similarly to the discussion
on C t in Eq. (4), C Z is dened as
CZ

ncC L sin b  C D cos bf1 ro=U x 2 g


.
2pr

Therefore, the total torque coefcient generated by the blades can be expressed as
Z r0
Z r0
T
1
1
C trq 1
dT

C Z K 2 2pr2 dr,

2
2
1
AD
rU
AD
rU
AD
r
r
h
h
0
0
2
2

(6)

(7)

ARTICLE IN PRESS
K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

958

Fig. 6. Overview of disk-loading method.

0.06

1.5
CD
CL

CD

CL

0.04
0.5
0.02
0
(Symbols: original data)
0.5
5

0
0
5
Angle of attack (deg)

10

Fig. 7. Drag and lift coefcients for relative angle of attack (a).

where A pD=22 and K U 2 =U 0 is the acceleration factor. The power coefcient is


estimated as follows:
To
o
1
3
3
2 rU 0 A
2 rU 0 A

Cw 1

r0

dT
rh

1
A

r0 

rh


ro
C Z K 3 2pr dr.
Ux

(8)

Such being the case, it is of interest that the performance of a wind turbine can be virtually
evaluated with Eqs. (7) and (8), though no swirl effect is incorporated into the calculation.

ARTICLE IN PRESS
K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

959

It will be later shown that this evaluation procedure provides a powerful tool to obtain
useful information of ow elds at low computational cost.
As shown in Fig. 5(b), concerning the inlet boundary condition, a uniform ow (U 0 ) was
specied. At the outlet boundary, zero-streamwise gradients were prescribed. Concerning the
top boundary, slip conditions were adopted. No-slip conditions were specied at the walls,
where the nearest node was properly placed inside the viscous sublayer. The number of grid
points in this study was 156  86. The Reynolds number (Re U 0 D=n) was set to 20000. It is
noted that the grid dependency and the Reynolds-number dependency of the computational
results were carefully examined in the previous report [11], where it was shown that the grid
resolution used was enough to obtain almost grid-independent solutions and also that the
relatively lower Reynolds-number condition did not cause any serious problems in
investigating fundamental ow characteristics, at least, for this kind of ow cases. This leads
to a practical way to reduce the computational cost (see for detail [11]).
Calculations were performed with the nite-volume procedure STREAM of Lien and
Leschziner [14], followed by several improvements and substantially upgraded by Apsley
and Leschziner [15]. This method uses collocated storage on a non-orthogonal grid and all
variables are approximated on cell faces by the UMIST scheme [16], a TVD
implementation of the QUICK scheme. The solution algorithm is SIMPLE, with a
RhieChow interpolation for pressure. In this study, to predict complex turbulent ow
elds with massive ow separation, the non-linear Eddy-viscosity model proposed by Abe
et al. [11] was incorporated into the code. This model is based on former models by Abe et
al. [1719], with some modications for the length-scale equation. Detailed descriptions of
the model are given in the reference paper [11].
4. Results and discussion
Fig. 8 compares the distributions of torque coefcient (C trq ) and power coefcient (C w )
between the bare and diffuser-shrouded wind turbines. In the gure, l is a tip speed ratio
dened as
r0 o
l
.
(9)
U0
It is found that the shapes of both C w and C trq are similar between the bare and diffusershrouded wind turbines. However, the diffuser-shrouded wind turbine returns much higher
output power compared to the bare wind turbine. The power coefcient obtained from the
diffuser-shrouded wind turbine is about four times as high as that from the bare wind
turbine. The basic performance of such a wind turbine has been discussed by many
research groups (see for example, Rauh and Seelert [3]; Hansen et al. [5]; Inoue et al. [10]),
from which it has been pointed out that a diffuser-shrouded wind turbine has the
possibility of giving C w exceeding the Betz limit 16=27 owing to the effect of the
diffuser placed around the wind turbine.
On the other hand, Fig. 9 compares the distributions of torque coefcient (C trq2 ) and
power coefcient (C w2 ) normalized by the local mean velocity just behind the turbine
blades (U 2 , see Fig. 5(a)). They are dened as
T
C trq
2 ;
2
K
2 rU 2 AD

C trq2 1

To
Cw
3.
3
K
2 rU 2 A

C w2 1

(10)

ARTICLE IN PRESS
K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

960

(a)

(b)
Fig. 8. Basic performance of wind turbines: (a) bare wind turbine; (b) diffuser-shrouded wind turbine.

Fig. 9. Performance normalized by local mean velocity just behind turbine blades.

In the gure, l2 is alternative denition of a tip speed ratio dened as


l2

r0 o l
.
U2
K

(11)

ARTICLE IN PRESS
K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

961

These values (C trq2 , C w2 and l2 ) are used to investigate the performance of the wind
turbine under the condition using the local mean velocity just behind the blades. It is of
interest that both the bare and diffuser-shrouded wind turbines return almost the same
peak performance as seen in Fig. 9, though some discrepancy is seen in the high l2 region.
This may indicate that the present wind turbine operates similarly against the approaching
wind even with a anged diffuser. Having considered these facts, it is thought that the
augmentation of the power in the diffuser-shrouded wind turbine is mainly caused by the
acceleration of the approaching wind by the anged diffuser. This is supported by the
previous study of Hansen et al. [5], which evaluated the effect of a diffuser on the power
augmentation obtained from a wind turbine. Abe and Ohya [11] also investigated the
acceleration mechanism of the approaching wind when the anged diffuser is placed
around a wind turbine.
Fig. 10 gives an overall view of the ow eld around the present anged diffuser at
l 4:5, in the form of the calculated streamfunction plots. As seen in the gure, a ange
generates a large separation behind it. This is a notable feature of this kind of ow eld,
which is different from that around a general airfoil-shaped diffuser with no ange. In fact,
this separation generates a low-pressure region, owing to which more wind can be drawn
into the diffuser [11].
Fig. 11 compares the computational results of C trq and C w with the experimental data. It
is found that the present calculation gives satisfactory predictions for both the bare and
diffuser-shrouded wind turbines. Although some overprediction is seen in higher l region
of the diffuser-shrouded wind turbine, the peak performance is well predicted and the
difference between the bare and diffuser-shrouded wind turbines is sufciently reproduced.
Fig. 12 shows distributions of the angles in Fig. 6 at l 2:8 for the bare wind turbine.
For this condition, the wind turbine gives high performance and thus no separation is
expected to appear at the blade surface. As seen in the gure, the calculated relative angle
of attack (a) is less than 10 degrees at every position, being sufciently involved in the
range shown in Fig. 7. On the other hand, in the low l region, it is easily expected that a
may become very large beyond the range in Fig. 7. To investigate ow phenomena in
detail, the angular momentum was estimated by the streamwise and rotational velocities.
The proles for two conditions, i.e., l 2:8 and 1:7 for the bare wind turbine, are

Fig. 10. Streamlines of the ow eld (l 4:5).

ARTICLE IN PRESS
962

K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

(a)

(b)
Fig. 11. Comparison of basic performance: (a) torque coefcient; (b) power coefcient.

80
Inflow angle
Angle of blade setting

60
Angle (deg)

Relative angle of attack


40

20

0
0

0.2

0.4
r/D

Fig. 12. Distributions of inow angle (b), angle of blade setting (g) and relative angle of attack (a b  g) at
l 2:8 for the bare wind turbine.

ARTICLE IN PRESS
K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

963

Fig. 13. Radial distributions of angular momentum for the bare wind turbine.

compared in Fig. 13. It is readily seen that the computational results at l 2:8 correspond
well to the measured data. This means that the present calculation can provide a
reasonable distribution in the radial direction as well as a bulk performance for conditions
with no separation at the blade surface.
On the contrary, the measured data in Fig. 11 show that the performance decreases at
l 1:7. As will be shown later, under this condition, separations and their related large
vortex-like structures appear on the outer (tip) side. They occupy the region of 1=3 of the
blade radius and are highly three-dimensional. Fig. 13 denitely explains that these large
ow structures do not work effectively to generate the angular momentum (i.e. torque).
Generally, such complex ow phenomena with large ow structures can never be modeled
just by 2-D database with some simple corrections of 3-D effect, even if 2-D database of an
airfoil are available for high angles of attack including two-dimensional massive
separation. Unfortunately, at least now, there is no data (or model) available to represent
such complex ow phenomena. Such being the case, calculations in the low l region were
not carried out in this study. Although the present results are encouraging from the
engineering viewpoint, further investigations will be needed to resolve this issue.
Computational results of the acceleration factor K are compared with those of the
experiment in Fig. 14, where U 2 is calculated by averaging the streamwise velocity in the
section just behind the wind turbine (i.e. the load imposed by Eq. (4)). It is seen from the
gure that the present computation returns generally reasonable trends for the variation of
K against l, though the computation gives some underpredictions in the higher l region
especially for the diffuser-shrouded wind turbine. Fig. 15 compares the calculated
streamwise-velocity proles immediately behind the wind turbine with those of the
experiment. The conditions were l 2:8 (l2 3:1) for the bare wind turbine and l 4:5
(l2 3:2) for the diffuser-shrouded wind turbine respectively. In the gures, the plots of
the experimental data have been made by averaging the streamwise velocity in the
rotational (y) direction. As described earlier, the similar condition of l2 allows us to
compare the results under the condition of similar local mean velocity (i.e. local tip speed
ratio) just behind the blades. As seen in the gure, a similar trend is seen in the velocity
proles. The velocity slightly decreases in the region around r=D 0:4, whereas it increases

ARTICLE IN PRESS
964

K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

Fig. 14. Comparison of acceleration factor.

2.5
Experiment

Ux/U0

Calculation

1.5
1
0.5
0
0

0.2

0.4

0.6

0.4

0.6

r/D

(a)
2.5

Experiment

Ux/U0

Calculation

1.5
1
0.5
0
0

(b)

0.2
r/D

Fig. 15. Comparison of streamwise velocity (x 48 mm): (a) bare wind turbine; (b) diffuser-shrouded wind
turbine.

ARTICLE IN PRESS
K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

965

again at the end of the blade tip (around r=D 0:5). In particular, in the region close to the
diffuser wall (Fig. 15(b)), the ow is greatly accelerated due to the tip clearance. The
present computation successfully predicts such a trend of the velocity distribution in the r
direction, though some underprediction is seen especially in Fig. 15(b). Velocity vectors in
the xr plane are compared in Fig. 16, where vector plots of the experiment have been
made by averaging the measured velocities in the y direction. It is found that the computed
velocity vectors generally correspond well to those measured in the experiment, though
some discrepancy is seen particularly in the downstream region of the bare wind turbine.
Fig. 16(b) indicates that the present computation can provide useful information on the
deceleration mechanism inside the diffuser. As seen in the gure, there is no separation on

(a)

(b)
Fig. 16. Comparison of velocity vectors in xr direction: (a) bare wind turbine; (b) diffuser-shrouded wind
turbine.

ARTICLE IN PRESS
966

K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

the diffuser wall even with a large diffuser opening angle of f 12 . This notable fact is
reasonably predicted by the present computation. From these results, it can be said that the
present computation sufciently reproduces fundamental features of the ow elds of this
kind.
Such being the case, the present calculation generally returns reasonable results. It is,
however, seen that the approaching wind speed commonly tends to be a little
underpredicted compared to the measured data. Since this trend is seen even in the cases
of the bare wind turbine, it is thought that the main reason may lie in some insufciency in
the model of the load and/or some inuence of ignoring the swirl effect in the calculation.
In this respect, further efforts are needed to construct an advanced prediction model,
including a new accurate disk-loading model applicable to the low l range. Although there
still remains some margin for improvement, it is said that the present computational
procedure is very useful for the rst assessment of performance in developing new diffusershrouded wind turbines.
Fig. 17 shows a comparison of the measured velocity vectors just behind the blades for
three representative rotating frequencies, i.e., 2100, 1500 and 900 rpm. It should be noted
again that the present experiment adopts such conditions that local tip speed ratios (l2 ) can
be similar between the bare and diffuser-shrouded wind turbines at the same rotational
frequency. As seen in Fig. 9, the conditions of 2100 rpm (l2 4) and 1500 rpm (l2 3) are in
the region of a higher tip speed ratio, where the ow is thought to be smooth with no
massive separation on the blades. On the other hand, 900 rpm (l2 2) is in the region of a
lower tip speed ratio and then the results give us information of the ow structure in the
stall condition. In the cases of 2100 and 1500 rpm, the ow successfully rotates in the y
direction, which means that the blades effectively generate torque. In the region close to
the blade tip, some vortex structures are seen, which are thought to be the tip vortices. It is
of interest that similar patterns can be seen for both the bare and diffuser-shrouded wind
turbines, though the tip vortex of the diffuser-shrouded wind turbine looks stronger than
that of the bare wind turbine. On the other hand, in the case of 900 rpm, very large ow
structures are clearly seen. These structures occupy the region of 1=41=3 of the blade
radius from the blade tip. This kind of large structure is usually thought to be caused by
massive separations occurring at the blade surfaces. They prevent the blades from
generating torque as described in Fig. 13 and thus the performance of the wind turbine
decreases as shown in Figs. 8 and 9. From these gures, it is generally said that the ow
structures of the bare and diffuser-shrouded wind turbines are basically similar to each
other, at least, at the location just behind the turbine blades. This corresponds well to the
results of the performance normalized by the local mean velocity (U 2 ) shown in Fig. 9. In
the region close to the blade tip, however, some discrepancies are seen mainly due to the
effect of the diffuser wall (i.e. the tip clearance).
To compare the ow structures of the bare and diffuser-shrouded wind turbines in the
region downstream of the blades, Fig. 18 shows the experimental data of velocity vectors
and streamwise vorticity distributions at three locations in the streamwise direction, i.e.,
x 48 mm (x=D 0:12), 148 mm (0.37) and 248 mm (0.52). As mentioned above, it is
found again that the ow structures of the diffuser-shrouded wind turbine just behind the
turbine blades are generally similar to those of the bare wind turbine except for the region
close to the blade tip. On the other hand, it is of great interest that completely different
trends are seen in the downstream region. In particular, rapid destruction of the vortex
structure is seen in the diffuser-shrouded wind turbine, while similar vortex structures

ARTICLE IN PRESS
K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

0.5

0.5

-0.5

5 [m/s]

-0.5

-0.5

5 [m/s]

-0.5

0.5

2100rpm ( = 3.6, 2 = 3.8)

0.5

5 [m/s]

-0.5

0.5

5 [m/s]

-0.5

0.5

5 [m/s]

(a)

0.5

-0.5

0.5

900rpm ( = 1.7, 2 = 1.8)

1500rpm ( = 4.5, 2 = 3.3)

0.5

-0.5

0.5

-0.5

1500rpm ( = 2.8, 2 = 3.1)

-0.5

2100rpm ( = 6.3, 2 = 3.9)

0.5

-0.5

967

5 [m/s]

-0.5

(b)

0.5

900rpm ( = 2.7, 2 = 2.1)

Fig. 17. Comparison of velocity vectors behind turbine blades (x 48 mm): (a) bare wind turbine; (b) diffusershrouded wind turbine.

ARTICLE IN PRESS
K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

968

0.5
0.5

vorticity

-0.5
5 [m/s]

-0.5

20
16
12
8
4
0
-4
-8
-12
-16
-20 [1/s]

vorticity

-0.5

5 [m/s]

-0.5

0.5

x=48mm (x / D = 0.12)

0.5

x=48mm (x / D = 0.12)

0.5

0.5

vorticity

-0.5

5 [m/s]

-0.5

20
16
12
8
4
0
-4
-8
-12
-16
-20 [1/s]

vorticity

-0.5

5 [m/s]

-0.5

0
0.5
x=148mm (x / D = 0.37)

20
16
12
8
4
0
-4
-8
-12
-16
-20 [1/s]

0
0.5
x=148mm (x / D = 0.37)

0.5

0.5

vorticity

-0.5

5 [m/s]

-0.5

(a)

20
16
12
8
4
0
-4
-8
-12
-16
-20 [1/s]

0
0.5
x=248mm (x / D = 0.52)

20
16
12
8
4
0
-4
-8
-12
-16
-20 [1/s]

vorticity

-0.5

5 [m/s]

-0.5

(b)

20
16
12
8
4
0
-4
-8
-12
-16
-20 [1/s]

0
0.5
x=248mm (x / D = 0.52)

Fig. 18. Comparison of velocity vectors and vorticity distributions in ry planes at 1500 rpm: (a) bare wind
turbine (l 2:8; l2 3:1); (b) diffuser-shrouded wind turbine (l 4:5; l2 3:3).

ARTICLE IN PRESS
K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

969

clearly remain even in the far downstream region of the bare wind turbine (see Fig. 18,
x=D 0:52). This is also supported by the variation of the streamwise vorticity at the three
locations.
5. Concluding remarks
Experimental and numerical investigations were carried out for ow elds of a small
wind turbine with a anged diffuser (i.e. a diffuser-shrouded wind turbine). By processing
the data obtained, characteristic values of the ow elds were estimated and compared
with those for a bare wind turbine. The main conclusions derived from the study are as
follows:

The present diffuser-shrouded wind turbine provided much a higher power output
compared to the bare wind turbine. The power coefcient of the diffuser-shrouded wind
turbine was about four times as high as that of the bare wind turbine. However, when
the performance was normalized by the local mean velocity just behind the turbine
blades, both the bare and diffuser-shrouded wind turbines returned almost the same
peak performance. This may indicate that the wind turbine operates similarly against
the approaching wind even with a anged diffuser. It is also thought that the power
augmentation of this kind of wind turbine is mainly caused by the acceleration of the
approaching wind by a anged diffuser.
The present computational results generally showed reasonable agreement with the
corresponding experimental data in the range from moderate to high l, where no
separation at the blade surface was expected. On the other hand, in the low l range, the
present experiment elucidated that very large vortex-like structures were generated.
These structures occupied the region of 1=41=3 of the blade radius from the blade tip.
This may indicate that further efforts are needed to construct a new disk-loading model
representing such complex ow phenomena in the low l range. Although there still
remains some margin for improvement, it is said that the present computational
procedure is very useful for the rst assessment of performance in developing new
diffuser-shrouded wind turbines.
Flow structures of the diffuser-shrouded wind turbine just behind the blades were
generally similar to those of the bare wind turbine, though some discrepancy was seen in
the region close to the blade tip. On the other hand, completely different trends were
seen in the region downstream of the wind turbine. In particular, it is worth noting that
rapid destruction of the vortex structure was seen in the diffuser-shrouded wind turbine,
while similar vortex structures were clearly seen even in the far downstream region of
the bare wind turbine. This is thought to be a notable feature of this kind of diffusershrouded wind turbine.

Acknowledgements
This research was partially supported by the Ministry of Education, Culture, Sports,
Science and Technology, Japan (Grant-in-Aids for Scientic Research, No. 14205139, No.
15360450), the Ministry of Economy, Trade and Industry (Matching Fund), Sumitomo
Fund (Environment Protection Research), Harada Memorial Fund (Fluid Machinery
Research) and Kyushu University (Program and Project for Education and Research). The

ARTICLE IN PRESS
970

K. Abe et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 951970

authors wish to express their appreciation to Mr. M. Matsubara of Kyushu University,


Fukuoka, Japan for his technical help in the experiments. KA also wishes to express his
appreciation to Professor M.A. Leschziner of Imperial College of Science, Technology and
Medicine, London, UK for the support in using the STREAM code.
References
[1] O. Igra, Research and development for shrouded wind turbines, Energy Conversion Manage. 21 (1981)
1348.
[2] B.L. Gilbert, K.M. Foreman, Experiments with a diffuser-augmented model wind turbine, Trans. ASME,
J. Energy Res. Technol. 105 (1983) 4653.
[3] A. Rauh, W. Seelert, The Betz optimum efciency for windmills, Appl. Energy 17 (1984) 1523.
[4] D.G. Phillips, P.J. Richards, R.G.J. Flay, CFD modelling and the development of the diffuser augmented
wind turbine, Proceedings of the Comp. Wind Eng. 2000, Birmingham, 2000, pp. 189192.
[5] M.O.L. Hansen, N.N. Sorensen, R.G.J. Flay, Effect of placing a diffuser around a wind turbine, Wind
Energy 3 (2000) 207213.
[6] M. Nagai, K. Irabu, Momentum theory for diffuser augmented wind turbine, Trans. JSME 53-489 (1987)
15431547 (in Japanese).
[7] I. Ushiyama, Introduction of Wind Turbine, Sanseido Press, Tokyo, 1997, pp. 7784 (in Japanese).
[8] Y. Ohya, T. Karasudani, A. Sakurai, Development of high-performance wind turbine with brimmed diffuser,
J. Japan Soc. Aeronaut. Space Sci. 50 (2002) 477482 (in Japanese).
[9] Y. Ohya, T. Karasudani, A. Sakurai, M. Inoue, Development of high-performance wind turbine with a
brimmed-diffuser: Part 2, J. Japan Soc. Aeronaut. Space Sci. 52 (2004) 210213 (in Japanese).
[10] M. Inoue, A. Sakurai, Y. Ohya, A simple theory of wind turbine with a brimmed diffuser, Turbomachinery
30 (2002) 497502 (in Japanese).
[11] K. Abe, Y. Ohya, An investigation of ow elds around anged diffusers using CFD, J. Wind Eng. Indust.
Aerodyn. 92 (2004) 315330.
[12] J.N. Sorensen, C.W. Kock, A model for unsteady rotor aerodynamics, J. Wind Eng. Indust. Aerodyn. 58
(1995) 259275.
[13] H. Kume, Y. Ohya, T. Karasudani, K. Watanabe, Design of a shrouded wind turbine with brimmed diffuser
using CFD, Proceedings of the Annual Conference of JSAS, The West Side Division, 2003, pp. 5154 (in
Japanese).
[14] F.S. Lien, M.A. Leschziner, A general non-orthogonal collocated nite volume algorithm for turbulent ow
at all speeds incorporating second-moment turbulence-transport closure, Part 1: Computational
implementation, Comput. Methods Appl. Mech. Eng. 114 (1994) 123148.
[15] D.D. Apsley, M.A. Leschziner, Advanced turbulence modelling of separated ow in a diffuser, Flow, Turbul.
Combust. 63 (1999) 81112.
[16] F.S. Lien, M.A. Leschziner, Upstream monotonic interpolation for scalar transport with application to
complex turbulent ows, Int. J. Numer. Methods Fluids 19 (1994) 527548.
[17] K. Abe, Y.J. Jang, M.A. Leschziner, An investigation of wall-anisotropy expressions and length-scale
equations for non-linear Eddy-viscosity models, Int. J. Heat Fluid Flow 24 (2003) 181198.
[18] K. Abe, T. Kondoh, Y. Nagano, On Reynolds stress expressions and near-wall scaling parameters for
predicting wall and homogeneous turbulent shear ows, Int. J. Heat Fluid Flow 18 (1997) 266282.
[19] K. Abe, T. Kondoh, Y. Nagano, A new turbulence model for predicting uid ow and heat transfer in
separating and reattaching owsI. Flow eld calculations, Int. J. Heat Mass Transfer 37 (1994) 139151.

You might also like