Renewable Energy 112 (2017) 25e34

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Aerodynamic interaction of diffuser augmented wind turbines in

multi-rotor systems
€ ltenbott a, *, Yuji Ohya b, Shigeo Yoshida b, Peter Jamieson c
Uli Go
a
Department of Aeronautics and Astronautics, Kyushu University, 744 Motooka, Nishi-ku, Fukuoka 819-0395, Japan
b
Research Institute for Applied Mechanics, Kyushu University, 6-1 Kasuga-koen, Kasuga 816-8580, Japan
c
Wind Energy DCT, University of Strathclyde, Royal Collage R336, Glasgow, G1 1XW, UK

a r t i c l e i n f o a b s t r a c t

Article history: The most common wind turbine is the singleerotor, horizontal axis wind turbine. In order to reduce the
Received 11 August 2016 cost of energy, upscaling of singleerotor wind turbines has been a major trend. Recent studies however
Received in revised form show that for a given technology, the cost usually rises when upscaling, notably due to increased masses.
18 January 2017
To reach capacities beyond 10 MW, multierotor systems (MRS) have promising advantages over single
Accepted 2 May 2017
Available online 8 May 2017
erotor systems (SRS). On the other hand, diffuser augmented wind turbines (DAWTs) can significantly
increase the performance of the turbine. In this research, brimmed DAWTs are introduced in a MRS. In
wind tunnel experiments, the aerodynamics of two and three DAWTs, spaced in close vicinity in the
Keywords:
Multierotor systems
same plane normal to a uniform flow, have been analyzed. Power increases of up to 5% and 9% for the two
Clustered wind turbines and three rotor configurations are respectively achieved in comparison to a singleerotor turbine. Hot
Diffuser augmented wind turbines ewire techniques used to measure the flow speed near the gap between the DAWTs in a MRS have
Wind turbine wakes shown an acceleration of the flow. Phenomena of bluff body flows are reviewed to analyze the physical
Upscaling of wind turbines dynamics of the flows in the MRS on the basis of the flow dynamics observed in a SRS.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction (MRS), defined as containing more than one rotor in a single

The most commonly used wind turbine for electricity genera-
tion is the horizontal axis, singleerotor wind turbine. A major trend
is to increase the size of the rotor in order to generate more energy
and to reduce the cost of energy generated from the wind. This was
notably possible due to technological advancements in fiber re-
inforcements to manufacture longer blades. Currently, researchers
focus on developing technologies to upscale singleerotor wind
turbines to 10 MW and beyond. However, there are certain issues
related to upscaling wind turbines which have been discussed by
Jamieson and Branney [1] and Sieros et al. [2] who stated that
upscaling usually results in an unfavorable weight increase for a
given technology level. On the other hand, Hofmann and Sperstad
[3] pointed out that low failure rates and short maintenance du-
rations are crucial to achieve low cost of energy. A concept to
overcome these issues for wind turbines is the multierotor system
* Corresponding author.
€ltenbott), ohya@riam.kyushu-
E-mail addresses: uli@riam.kyushu-u.ac.jp (U. Go
u.ac.jp (Y. Ohya), yoshidas@riam.kyushu-u.ac.jp (S. Yoshida), peter.jamieson@
strath.ac.uk (P. Jamieson).
structure. Upscaling by increasing the number of rotors instead of
the diameter of a single rotor leads to significant savings in mass. A
reduction in mass consequently reduces the overall cost of the
system.
Proposed as early as 1875 in Denmark [4] and later in the 1930s
by Honnef in Germany, as reported by Jamieson [5], the multi-rotor
concept has been further developed by Heronemus [6] in the 1970s,
while Smulders et al. [7] later analyzed a two rotor configuration in
a wind tunnel. In 2010, the South-West Research Institute con-
ducted an analysis of a 7 rotor configuration in the NASA Langley
wind tunnel to test the concept of Heronemus and found no
negative interferences between the rotors [8]. Simulations of the
aerodynamic performance for the same 7 rotor system were pre-
sented by Chasapogiannis et al. [9]. In their CFD simulations using
actuator discs, a power increase of 3% was predicted for the 7 rotor
system. On a larger scale, structural considerations of a 20 MW
multi-rotor system with 45 rotors have been presented by Jamieson
and Branney [1]. They showed that mass and cost are reduced, in
comparison to a single-rotor turbine with the same power, and that
the structure is designed based on the aerodynamic forces rather
than on the rotor loading.

http://dx.doi.org/10.1016/j.renene.2017.05.014
0960-1481/© 2017 Elsevier Ltd. All rights reserved.
26 €ltenbott et al. / Renewable Energy 112 (2017) 25e34
U. Go
The effect of closely spaced wind turbines in a lateral wind farm
configuration was investigated by McTavish et al. [10] and Meyer
Forsting et al. [11]. Close proximity of the rotors showed beneficial
effects on the performance of three or more wind turbines due to
an in-field blockage effect. The average power increase was be-
tween 5 and 6% of three conventional wind turbines at a lateral
spacing with 0.5 rotor diameters separation between the tips of
neighboring rotors.
Innovations to increase the performance of wind turbines have
led to the use of ducts around the rotor to augment the flow, these
turbines are commonly called diffuser augmented wind turbines
(DAWTs). Extensive experimental analysis have been performed by
many researchers on DAWTs over the past 50 years, see for example
Rainbird and Lilley [12], Foreman et al. [13], Igra [14] and Abe et al.
[15], to name a few. In 2007, summarizing the experimental data
and comparing them with CFD results by Hansen et al. [16] and
momentum theory, van Bussel [17] concluded that the performance
increase of DAWTs is proportional to the mass flow through the
duct and that larger performances are possible by lowering back
pressure levels at the diffuser exit. To achieve lower back pressure, a
brim was attached to the diffuser exit by Ohya and Karasudani [18].
The brim plays an important role in generating vortices resulting in
the reduced back pressure. For three brimmed DAWTs, rotor per-
formance increases of 2e5 fold have been achieved in wind tunnel
experiments [18]. However, although they promise higher perfor-
mances, DAWTs are still not widely used. Further, there are struc-
tural challenges in upscaling diffusers to be used with large wind
turbines.
Research presented by Wang and Chen [19] analyzed two
DAWTs which yawed individually with wind from different angles.
They have placed the turbines side-by-side with gaps of s/D of 0.22,
0.44 and 0.67, where s is the gap width and D is the external
diameter of the DAWT. The focus of the study was to investigate the
deficit of a DAWT in the wake of a turbine upstream at different
wind directions. Although not discussed in the paper, the graphs
show a small increase in power, of approximately 2e5%, when the
DAWTs are set normal to the inflow.
To understand the fundamental phenomena of a MRS with
brimmed DAWTs, Ohya et al. [20] conducted experiments
comparing drag coefficients of circular flat disks with drag and
power coefficients of brimmed DAWTs in side-by-side configura-
tions. Three circular flat disks showed large differences in the drag
coefficients of each disk. The differences in drag coefficients
strongly correlated to biased gap flow phenomena measured with a
hot wire technique in the near wake. Although weaker in magni-
tude, similar variations in drag coefficients of three DAWTs have
been observed. It was shown that the drag coefficients of DAWTs
correlate well with their power coefficients. This lead to the
conclusion that differences in power coefficients of DAWTs in a
MRS configuration are affected by gap flow patterns.
In wind tunnel experiments presented in this paper, the effects
of the spacing of brimmed DAWTs with various brim heights on
their aerodynamic performance are analyzed. The turbines are set
in the same plane, next to each other, without the ability to yaw
individually since a multi-rotor system is meant to yaw as a whole.
Two and three turbines are placed in side-by-side configurations to
measure power coefficients. The average power coefficients of the
MRS configurations are compared with a SRS. The analysis has been
carried out for MRS using conventional turbines and DAWTs.
Fundamental flow phenomena from bluff bodies, including vortex
shedding, are reviewed to describe the results obtained from the
experiments. The phenomena include the flow around multiple
bluff bodies and vortex shedding from three dimensional bodies.
The physical dynamics of the flows in the MRS are analyzed on the
basis of the flow dynamics observed in a SRS.
2. Experimental setup
The wind turbines are placed in the wind tunnel test section,
which has a cross section of 3.6 m by 2 m and a length of 15 m. An
outline of the boundary layer wind tunnel can be seen in Fig. 1 [21].
The test section consists of 6 elements, where the side walls and
ceiling plates have been removed for the elements number 3, 4 and
5. With the removal of the side walls and ceiling plates, the wind
tunnel becomes semi-open, which significantly decreases blockage
effects.
The rotor plane of the turbines is positioned 1 m downstream
from the first opened side walls. Configurations of two turbines and
three turbines, placed side-by-side normal to the uniform incoming
flow, are analyzed. The setup for the three DAWTs MRS configu-
ration can be seen in Fig. 2.
The rotor shafts of the turbines are located 0.65 m above the
floor, which leaves a distance of approximately 0.31 m between the
brim and the floor. The effects from the ground effect on the torque
of the turbines has been evaluated and showed minor difference a
distances from 0.29 m to 0.62 m between brim edge and the
ground. Horizontally, the turbines are aligned with the center of the
wind tunnel. The minimum distance to the opened side walls is
0.4 m. The effects of the vicinity of the turbines to the side walls,
floor and ceiling on the results have been tested and showed
negligible small variations. The shape of the diffusers, with the
attached brims, can be seen in Fig. 3(a).
All diffusers used in this study have the same diffuser shape
denoted as Ci-Diffuser. The stream wise depth, Lt, of the Ci-Diffuser
is 0.137 throat diameter, Dthroat, defined as the inside diameter of
the diffuser, and follows a cycloid curve. The brims have heights
from 0.03 to 0.1 Dthroat (3e10%). The denotation used is B03 for 3%,
B05 for 5%, B07 for 7% and B10 for 10%, respectively [16]. Brim
height, h, projected height, H, and diffuser depth, Lt, are illustrated
in Fig. 3(b).
2.1. Power measurements
The rotors are driven by a motor using a torque meter as a link
between the rotor shaft and the motor shaft, as shown in Fig. 4.
The rotational speed u is set constant and torque t is measured
with the signal fed into the PC through an AD-Converter. Sampling
time for the measurements is 20 s at a sampling frequency of
100 Hz. Measurement error is estimated to be in the order of ±0.5%.
To adjust the gap, s, between the turbines, the turbines are
Fig. 1. Wind tunnel specifications with side walls and ceiling plates removed for ele-
ments number 3, 4 and 5.
 €ltenbott et al. / Renewable Energy 112 (2017) 25e34
U. Go
Fig. 2. Turbine setup in the wind tunnel, with view in the flow direction showing the
three DAWTs MRS configuration.
moved horizontally on a rail. Approximately 1 Drotor downstream,
where Drotor is the external diameter of the rotor of one turbine, a
torque meter is connected to the rotor shaft. On the other end of the
torque meter, a speed-controlled motor is connected. Assuming
constant inflow velocity, the optimum tip speed ratio is realized by
setting a constant rotational speed of the rotor. For the period of
measurement, the motor speed is held constant by the motor
controller. The two turbine configuration, with a description of the
major dimensions, is illustrated in Fig. 5.
The rotor blades, produced with a 3D-printer, have cross-
sections of MEL-type airfoils [22]. At a wind speed of 7 m/s, the
Reynolds numbers Re are shown in Table 1 using the external
diameter D of the turbines as a characteristic length L.
For the conventional wind turbine and the DAWT, D is the rotor
diameter and the brim diameter, respectively. The Reynolds num-
ber regarding the airfoil uses the chord length of the blades as a
characteristic length L. The Reynolds number is obtained using:
U0 $L
Re ¼ ; (1)
n
where U0 is the wind speed of the approaching flow in the wind
tunnel and n is the kinematic viscosity of the air at the ambient
temperature of the experiment (n ¼ 1.511  105 m2/s). The Rey-
nolds number at the tip of the blade is in the order of 105. The
(a)
Fig. 3. (a) Shape of diffuser with brims attached at the diffuser exit normalized to throat diameter, Dthroat for Ci-Diffuser with brim heights from 3 to 10%. (b) Dimensions of the
brimmed diffuser.
27
Fig. 4. Side view of the measurement setup.
rotation direction of the rotors is clockwise for all turbines, as seen
in the flow direction.
2.2. Hot-wire measurements
A hot wire technique is used to measure the flow velocity
component in the streamwise direction. The single wire probe,
perpendicular to the streamwise flow is mounted on a rod which is
attached to a traverse system. The hot wire measures the flow ve-
locity on the horizontal center line between the rotor shafts of the
turbines in increments between 5 and 10 mm. The measurements
are taken in the wake at a distance of 0.08 Dthroat behind the brim,
denoted as b in Fig. 5. A sampling time of 10 s, with a sampling
frequency of 1 kHz is used to collect the flow velocity measure-
ments in the wake of the turbines. 1 kHz sampling frequency is well
above the occurring frequencies of the rotor rotational speed
(14e18 Hz) and vortex shedding frequencies (18 Hz) at the brims.
2.3. Blockage in the wind tunnel
Placing obstacles in the wind tunnel causes blockage effects.
McTavish et al. [10] reported no impact on rotor performance with
blockage ratios lower than 10% in a closed wind tunnel test section
with three conventional wind turbines. The area of the cross sec-
tion in the wind tunnel is 7.2 m2. Table 2 shows the blockage ratio
for the various configurations studied in this work.
It can be seen that 10% blockage is exceeded when using three
DAWTs, where the maximum blockage reaches almost 15%. By
removing side walls and ceiling plates in the test section, the
blockage effect can be reduced significantly. This was demonstrated
in an experiment [18] with a DAWT of nearly 1.5 m external
diameter. With a blockage ratio of approximately 24% using the
same semi-open wind tunnel as the one in this work, the experi-
mental results agree very well with data from field experiments.
This confirms the assumption that any effect of blockage is reduced
(b)
28 €ltenbott et al. / Renewable Energy 112 (2017) 25e34
U. Go

Fig. 5. Schematics of two brimmed DAWTs placed side-by-side (top view).

Table 1
Characteristic length L and corresponding Reynolds numbers Re.

Conventional CiB03 CiB05 CiB07 CiB10 Blade (L ¼ chord length)

L ¼ D [m] 0.512 0.595 0.615 0.636 0.667 0.06

Re  105 2.37 2.76 2.85 2.95 3.09 0.28

Table 2
Blockage ratio of the wind tunnel for the configurations studied.

Conventional Drotor ¼ 0.512 m CiB03 D ¼ 0.595 m CiB05 D ¼ 0.615 m CiB07 D ¼ 0.636 m CiB10 D ¼ 0.667 m

AT [m2] 0.206 0.278 0.297 0.318 0.349

1 AT/AWT 2.86% 3.86% 4.13% 4.41% 4.85%
2 AT/AWT 5.72% 7.72% 8.25% 8.82% 9.71%
3 AT/AWT 8.58% 11.59% 12.38% 13.24% 14.56%

to negligible levels.
3. Bluff body flow
Studying the flow around bluff bodies of similar shape as the
brimmed DAWTs has been very useful to understand the basic flow
phenomena. Relevant to the flow around multiple rotors sur-
rounded by brimmed diffusers, two bluff body flow phenomena are
reviewed. Firstly, the flow around multiple bluff bodies is discussed
for flat plates and circular cylinders. Secondly, the behavior of
vortex shedding from a square prism is explained. These flow
phenomena and their applications to wind turbines have been
summarized in details by Ohya [23].
3.1. Flow around multiple two dimensional bodies
The flow through the gap of a pair of flat plates arranged side-
by-side perpendicular to the inflow showed that biasing of the
gap flow in the near wake influences the drag coefficients of each
plate [24]. The flat plate on the side of which the flow is biased
shows higher drag and regular vortex shedding, where the plate on
the other side shows the opposite. The bi-stable flow patterns
occurred in a stable manner for a certain time period until it
changed to the other pattern. The biased gap flow was observed for
gap ratios s/D < 2, where s is the gap width and D is the body width.
An example of possible biased gap flows is illustrated in Fig. 6 (a) for
the flow through the gap of two flat plates. An example of possible
biased gap flows is illustrated in Fig. 6(a) for the flow through the
gap of two flat plates.
A pair of circular cylinders, positioned perpendicular to the
inflow also shows biased gap flow patterns for gap ratios s/D < 1 in
a similar way as the pair of flat plates. However, the gap flow pat-
terns through the circular cylinders shifted randomly at irregular
intervals [24]. The flow direction is illustrated in Fig. 6 (b).
A DAWT with air flowing through the rotor can be regarded as a
solid ring structure with free flow around the outside and restricted
flow through the inside where the rotor is located. Assuming the
DAWT acts as a circular flat disc, the body width is the brim
diameter D and the gap ratio is taken as s/D. With air passing
through the inside of the diffuser, the projected width of the ring
structure H can be regarded as a second parameter for the gap ratio
s/H. In Table 3, the gap ratios s/D and s/H are shown for each
experimental setup with all diffuser-brim combinations.
3.2. Vortex shedding from three dimensional bodies
Nakamura and Ohya [25] used two hot wires to estimate the
vortex shedding frequency on two sides of a square prism. They
found that the vortex shedding occurs either in one or the other
plane. However, the vortex shedding around the circular brim of a
DAWT appears to be random [23]. With the introduction of so-
called vortex control plates, the shedding around the circular
circumference of the brim can be organized and strengthened in
each segment, resulting in small increases in performance, as re-
ported by Ohya [23].
When placing two wind turbines side-by-side, the general
appearance of the setup can be seen as one body being much wider
than its height. Therefore, vortex shedding is likely to happen in the
plane of the shorter distance. In a side-by-side configuration of two
or three DAWTs, the vortex shedding preferably occurs over the
edges of the brim facing the top and the bottom.
4. Measurement procedure for turbine power output
First the power coefficient of each turbine as a SRS is individu-
ally measured. The power coefficient for this single turbine
configuration, CP0i, is calculated as:
 €ltenbott et al. / Renewable Energy 112 (2017) 25e34
U. Go 29

(a) (b)

Fig. 6. (a) Gap flow patterns between two flat plates (top view); (b) Flow through the gap of two circular cylinders (top view).

Table 3 of the DAWTs are significantly above the conventional turbine.

Ratio s/H of gap s to projected width of the duct H, as a function of the gap ratio s/D Within the diffuser augmented turbines, the best performance is
for the configurations studied.
achieved with brim heights between B05 and B10.
s/D The maximum power coefficient, CP0i of each turbine in SRS
0 0.05 0.1 0.15 0.2 0.3 0.5 configurations is measured at their optimum tip speed ratio. The
average of these maximum power coefficients of all n turbines in
s/H CiB03 0.00 0.75 1.51 2.26 3.01 4.52 7.53
CiB05 0.00 0.62 1.24 1.86 2.48 3.73 6.21 SRS configurations is defined as:
CiB07 0.00 0.53 1.06 1.59 2.12 3.18 5.30
CiB10 0.00 0.44 0.88 1.33 1.77 2.65 4.42 CP01 þ / þ CP0n
CP0 ¼ (3)
n
After that, the power coefficients of the wind turbines in two
and three turbine configurations are measured simultaneously. In
PTurbine t$u the multi-rotor configurations, the power coefficient is denoted as
CP0i ¼ ¼1 (2)
2 rAU0
PWind 3 CPi for each turbine. The average power coefficient of the MRS
containing n turbines is then defined as:
with the power in the wind, PWind, calculated with the air density, r,
the wind speed, U0, and the area, A, swept by the rotor. The rotor CP1 þ / þ CPn
CP ¼ (4)
swept area, A, is used as a reference for both the conventional n
turbine and the DAWT in order to compare purely rotor perfor- To understand how the power coefficient changed in each tur-
mance. For its part, the power from the turbines, PTurbine, is calcu- bine, the CPi from each turbine in the MRS configurations is
lated from torque, t, measured by the torque meter and the angular compared to the CP0i of each in the SRS configurations. Therefore,
velocity, u, of the rotor (Fig. 4). Fig. 7 shows the power curves, as a the variation, DCPi is defined as:
function of the tip speed ratio, l, of a conventional turbine and

DAWTs using a Ci-Diffuser with brim heights ranging from B03 to CPi
DCPi ¼ 1 (5)
B10. CP0i
At l ¼ 4.6, the DAWT with B10 shows an irregularity which is
In order to compare the average power coefficient of the MRS
assumed to be a measurement error that originates from a reso-
with the average power coefficient of the wind turbines as SRS, the
nance frequency of that rotational speed of the rotor. Therefore, the
variation, DCP is defined as:
wind speed, U0, was chosen to make sure that the peak of
maximum power coefficient is not affected by errors from reso- !
nance frequencies. The results clearly show that the performances CP
DCP ¼ 1 (6)
CP0

In the MRS configuration DCP is estimated for each gap ratio s/D
from 0 to 0.5 for the two turbine MRS and s/D from 0 to 0.2 or 0.3 for
the three turbine MRS. It should be mentioned that the optimum
tip speed ratio didn’t change significantly when clustering the wind
turbines in a MRS.

5. Performance of multiple turbines

The performance of multiple turbines in an array was evaluated

by placing two and later three turbines side-by-side and by
measuring the power output in the wind tunnel.

5.1. Average power output of turbines

First, the power output of a MRS with two turbines placed side-
Fig. 7. Power coefficients as a function of the tip speed ratio of a conventional turbine by-side was measured and compared to each turbine measured in a
and DAWTs in SRS. SRS configuration. The smallest gap for the two conventional
30 €ltenbott et al. / Renewable Energy 112 (2017) 25e34
U. Go

turbines was chosen to be 2.5% of the rotor diameter, mimicking a

safety gap to avoid collision of the rotor blades. The turbines in the
MRS configurations show a power increase of approximately 1% for
all the gap ratios ranging between 0.025 and 0.5 of the external
diameter D (in the case of the conventional turbine, the external
diameter D is the rotor diameter Drotor) as seen in Fig. 8.
For their part, at s/D ¼ 0, the DAWTs only exhibit small increases
in power. In this setup, the brims of neighboring turbines are in
direct contact. When further increasing the gap between turbines,
the power increases to a maximum of 3% for CiB03 and 5% for CiB10,
as seen in Fig. 8. Maximum power increases were measured be-
tween s/D ¼ 0.3e0.5. If and how much the CP will continue to in-
crease is unknown. In an experiment with two DAWTs with a rotor
diameter of 0.3 m, we observed that interferences beyond a gap
ratio of s/D ¼ 1 diminish and consequently CP drops back to the Fig. 9. Average power variation as a function of the gap between three turbines side-
by-side in a MRS. Setup of three DAWTs, CiB10 with gap of 0.15 D in the wind tunnel
values of a SRS. Also, the smaller the brim height, the earlier CP
(inset photo).
returns to the level of a SRS.
Next, the power coefficients of a MRS with three turbines side-
by-side are measured at various gaps. Fig. 9 shows that a MRS with
conventional turbines has an increase in power output between 3
and 5%. The source of oscillation of the measured CP is yet
unknown.
For the DAWTs, an increase at 0 D gap starting from 2% with
CiB05 to almost 5% with CiB10 was observed. The peak of power
increase was reached at a gap of 0.15 D in most cases. Maximum
power increase reached 4.5% for the conventional turbines and 9%
for the DAWT using the CiB10 configuration. The three turbine
configuration was measured until gap ratios of s/D ¼ 0.2 and 0.3 in
order to stay well inside the test section of the wind tunnel.

5.2. Individual turbine power output

Fig. 10. Individual turbine power variation of a MRS with three DAWTs in the CiB05
configuration.
For the three turbine MRS configurations, when looking at the
power output of each individual turbine, there are cases where a
turbine is not equally contributing to the trend of the average po-
wer output of the three turbine MRS. Fig. 10 shows the increase in
power output of the three DAWTs in the CiB05 configuration.
The power increase for turbines on the left, middle and right
side is shown with the average power increase. It can be observed
that the turbine on the left side shows very similar power increase
as the average power increase. The turbine in the middle shows
higher increases and the turbine on the right side shows lower
increases than the average. The biggest difference in power in-
crease was approximately 5% observed between the center turbine
and the turbine on the right at a gap ratio s/D ¼ 0.1. Fig. 11 shows
the increase in the power output of the three DAWTs in the CiB10
configuration. In this case the difference in power output is rather Fig. 11. Individual turbine power variation of a MRS with three DAWTs in the CiB10
negligible. configuration.
Differences in power increase of three conventional turbines in a

MRS configuration were also observed. Between the three con-

ventional wind turbines, the difference in power output was below
3%.

5.3. Hot-wire measurements in the near wake

In separate measurements, the flow velocity in the near wake is

Fig. 8. Average power variation as a function of the gap between two turbines side-by-
side in a MRS. Setup of two DAWTs, CiB05 with gap of 0.15 D in the wind tunnel (inset
photo).
measured by a hot wire technique at a distance of b ¼ 0.08 Dthroat
behind the brim, which is 0.1 Dthroat behind the rotor (as shown in
Fig. 5). Fig. 12 shows measurements for two DAWT’s with CiB10 at a
gap of s/D ¼ 0.15.
The red and blue dotted lines represent each DAWT as individual
SRS configurations, while the black dotted line shows the wind
speeds in the two DAWTs side-by-side in the MRS configuration
 €ltenbott et al. / Renewable Energy 112 (2017) 25e34
U. Go 31

Fig. 12. Two DAWTs (CiB10) in MRS side-by-side configuration (MRS-2-sbs) with a gap of 0.15 D. Hot wire measurements at b ¼ 0.08 Dthroat behind the brim and comparison with
each of the two DAWTs positioned individually as SRS on the left and right side of the wind tunnel.

(MRS-2-sbs). In this MRS configuration, the difference in power
increase between the two turbines was 4.1%, with a higher power
increase for the turbine on the left.
From the red and blue dotted lines of the SRS turbines, it can be
seen that the approaching flow accelerates by approximately 15% at
the edge of the brim. In the two turbine MRS configuration, the flow
in the gap shows an acceleration of up to 25% compared to the free
stream velocity. The edges of the brim facing outwards showed the
same acceleration value as observed for the SRS. The flow velocity
in the gap was rather symmetric, thus not indicating a biased gap
flow. The flow velocity has also been measured between two
DAWTs with CiB10 further downstream, at a distance of 0.3 and 0.5
Dthroat from the wind turbine rotors. At gap ratios s/D ¼ 0.1 and s/
D ¼ 0.2, no asymmetries were observed in the velocity profiles at
those positions further downstream.
Another feature that can be observed in the two turbine
configuration is the accelerated flow behind the rotors on the side
towards the gap. However, the signal from the hot wire measured
behind the rotor showed high fluctuations and thus contains
uncertainties.
For the three turbine MRS configuration, using CiB10 diffusers at
a gap of s/D ¼ 0.1, the flow speed is shown as the black dotted line in
Fig. 13.
The green, red and blue dotted lines show the velocity profiles of
each individual turbine in the SRS configuration. As previously seen
in Fig. 11, each turbine showed almost the same power increase,
with only small variations.
The flow velocity through each of the two gaps between the
three turbines accelerated by approximately 30%. The acceleration
at the outside showed the same acceleration of approximately 15%,
compared to the flows observed in the SRS configurations. There
was no asymmetry observed in the velocity profiles for the flow in
the configuration with the gap of 0.1 D.
6. Discussion
The experimental results show that the MRS configurations of
two conventional turbines don’t exhibit significant increase in po-
wer output at all gap ratios between s/D ¼ 0.025e0.5. These find-
ings confirm what has previously been reported for conventional
turbines with narrow gaps [7,8,10]. For MRS configurations of three
conventional wind turbines, a more notable power increase of
almost 5% was observed at gap ratios between s/D ¼ 0.025e0.3.
Apart from oscillating power output between s/D ¼ 0.025e0.15, the
power output seems independent from the gap ratios. Research by
McTavish et al. [10] has shown that the power increase originates
from an in-field blockage effect at a gap ratio of s/D ¼ 0.5, which
was characterized by accelerated flow through the gap between the
rotors.
Using DAWTs in two or three turbine MRS configurations

Fig. 13. Three DAWTs (CiB10) in MRS side-by-side configuration (MRS-3-sbs) with a gap of 0.1 D. Hot wire measurements at b ¼ 0.08 Dthroat behind the brim and comparison with
each of the three DAWTs positioned at the left, center and right side of the wind tunnel.
32 €ltenbott et al. / Renewable Energy 112 (2017) 25e34
U. Go
showed increased power output, which is a function of the gap
ratios. The maximum increase in average power output from two
DAWTs was between 3 and 5%. For three DAWTs, the maximum
average power increase was between 5 and 9%.
The power increase of each wind turbine in the MRS showed
differences compared to the average power increase of the MRS.
The differences of power increase between conventional wind
turbines in a MRS was 3% and for DAWTs in a MRS was 5%. Dif-
ferences in power increase between conventional turbines have
been reported previously by other researchers [10]; however, the
reasons for the differences haven’t been discussed in details.
Therefore, it is not known how the in-field blockage affects the
power output of each wind turbine in a MRS.
To measure the flow velocity profiles in the near wake, a hot
wire technique was used at b ¼ 0.08 Dthroat behind the brims of
DAWTs with CiB10. The single-rotor DAWT showed flow accelera-
tion of approximately 15% behind the edges of the brim. When
placing two turbines at a gap of 0.15 D, the gap flow accelerated by
25% compared to the free stream. With three turbines at a gap of 0.1
D the gap flow accelerated by approximately 30% in both gaps. The
level of acceleration seems be mainly related to the gap width. In all
cases, no asymmetry in the flow velocity profile in the gap was
observed. The flow acceleration at the sides of the two and three
turbine MRS showed the same values as the SRS, thus confirming
that blockage effects from the wind tunnel can be neglected. The
power increase of DAWTs in a MRS can be attributed to the in-field
blockage, as indicated by the flow acceleration measured in the gap.
One of the aerodynamic mechanisms of the DAWT is the vortex
shedding from the brim that reduces the back pressure and
therefore increases the power output of the turbine [18]. When
turbines are placed next to each other with the brims in direct
contact, the vortex shedding is inhibited at that point which would
result in a decreased power output. However, small power in-
creases were observed for both two and three turbine MRS con-
figurations at a gap ratio s/D ¼ 0. As reported by Ohya [23], vortex
shedding around a brimmed DAWT occurs in a random manner
around the circumference of the brim. Research by Nakamura and
Ohya [25] showed that vortex shedding of a square prism occurs
randomly in one or the other plane. Two DAWTs placed directly in
contact side-by-side appear as one object being twice as wide as it
is high. Therefore, vortex shedding could be strengthened over the
shorter plane, thus maintaining or even decreasing the back pres-
sure, which leads to a power increase.
The acceleration of the flow in the gap, measured near the flow
separation point at the brim leads to stronger shedding. This phe-
nomenon would also contribute to decrease the back pressure and
hence increase the power output of the turbines.
For gap ratios s/D larger than 0, vortex shedding becomes
possible around the whole circumference of the brim edge of each
turbine’s brim, which contributes to higher power output of the
MRS using DAWTs.
For narrow gaps, each of the vortices shed from opposite brim
edges tend to engulf into one another resulting in a direction
change of the flow. If this biased gap flow is stable for a certain
period of time, contrary drag forces were measured at neighboring
bodies. Several flow patterns are possible, each being bi-stable and
therefore a change in pattern is possible at random intervals [24].
Similar to the drag forces with bluff bodies, the rotor perfor-
mance of DAWTs is affected by biased gap flows due to the reduced
back pressure. Depending on the gap flow pattern present at the
time, the power coefficient of each of the three DAWTs in the MRS
configuration can show differing values. In the case presented in
Fig. 10, using CiB05 at gap ratios s/D ¼ 0.1 and 0.15, the DAWT in the
center shows the highest increase in power similar to what has
been reported for conventional wind turbines due to the in-field
blockage. The differences between the DAWTs power increase on
the left and right could originate from a biased gap flow due to
vortex interaction, which explains why the power output of the
turbines is not symmetric.
The results from the three turbine configuration using CiB10
seen in Fig. 11 don’t show significant differences in power increase
between each turbine. The gap flow acceleration is the same in both
gaps, without indication of a biased gap flow. The in-field blockage,
which showed the highest power increase in the center wind tur-
bine, seems to be compensated by possibly a biased gap flow, which
results in a rather symmetric and evenly distributed increase in
power output of the turbines. Further research is needed to clarify
the effects of in-field blockage and gap flow behavior on the per-
formance of conventional and DAWTs.
In an attempt to visualize the biased gap, a flow smoke wire
technique was used in the CiB05 and CiB10 configurations. In order
to get clear smoke lines, a wind speed of U0 ¼ 3 m/s was used,
which results in a Reynolds number of less than half of the one used
in the power measurements. With such a low Reynolds number, the
flow around the blades isn’t as well established, which resulted in
unsuitable power measurements. Nonetheless, observations
showed small biasing of the flow, which randomly shifted direction
in irregular time intervals. It wasn’t possible to see a stable flow
pattern that could be linked to the differences in power output,
which were measured at higher Reynolds numbers. Therefore, no
additional evidence of biased gap flow in relation to differences in
power output could be measured at this point.
The flow around multiple DAWTs with all the interactions of
vortices shed from the brimmed diffuser and the rotating wake
behind the rotor comprise a rather complex flow. The flow accel-
erations measured with a hot wire technique in the near wake are
an indicator of the in-field blockage. However, this doesn’t seem to
be able to capture the flow characteristics of biased gap flows that
would lead to the differences in power output measured at the
rotor shafts. Therefore, a CFD analysis is suggested to shed more
light into the flow characteristics around brimmed DAWTs in a
MRS. Furthermore, the combination of side-by-side with vertical
spacing of wind turbine arrays could confirm some of the as-
sumptions given in this manuscript.
7. Conclusions
In the wind tunnel experiments, the aerodynamics of MRS
configurations of two and three DAWTs spaced side-by-side in
close vicinity in the same plane normal to a uniform flow have been
analyzed. For gap ratios s/D between 0 and 0.5, conventional tur-
bines and DAWTs showed an increase in power output. DAWTs
show higher power increases than conventional turbines, reaching
a maximum of 9% at a gap ratio, of s/D ¼ 0.15 between neighboring
brims. The power increase is further dependent on the brim height
of the diffuser. The highest power increase was observed with a 10%
brim height (CiB10). It can thus be concluded that DAWTs installed
in side-by-side MRS configurations, with suitable spacing and
appropriate brim heights, increase the efficiency of the same wind
turbines taken in isolation.
In some cases, each DAWT in a MRS showed differences in their
individual power output compared to the average power output of
the whole MRS. With flow phenomena from bluff bodies, these
differences can be explained. From the flow around multiple
bodies, biased gap flows were observed that affected the drag co-
efficients of each bluff body and the average of the bluff bodies [24].
Similarities between the drag coefficients of bluff bodies and power
coefficients of DAWTs have been shown.
Using a smoke wire technique, the gap flow at low wind speeds
didn’t show any distinctive pattern. For its part, the velocity profile
 €ltenbott et al. / Renewable Energy 112 (2017) 25e34
U. Go 33

closely behind the brims of the DAWTs has been measured by a hot
wire technique. At a distance of b ¼ 0.08 Dthroat behind the brim, the
measurements showed flow accelerations in the gaps. There was no
asymmetry in the velocity profile that would suggest a biased gap
flow. The magnitude of the acceleration seems to be mainly
dependent on the gap ratio.
The overall power increase of a MRS can be explained with the
in-field blockage, which is indicated by the accelerated flow
through the gaps.
Furthermore, the flow around square prisms has shown that
vortex shedding prefers to occur at opposite edges of the prism
[25]. This means that vortex shedding prefers to occur in one or the
other plane, but not in both planes at the same time. In a two or
three turbine MRS, the DAWTs appear as one object with a width
two or three times the height. Vortex shedding could be
strengthened due to the MRS appearance. This can be seen as the
reason why the power of DAWTs in direct contact still show an
increase in power output.
When the gap between DAWTs is big enough to enable vortex
shedding at each side of the brim, accelerated flow velocities have
been measured with hot wire. Higher flow velocities at the edge of
the brim, where the flow separation occurs, lead to stronger
vortices resulting in reduced back pressure, which in return in-
crease the performance of the brimmed DAWT in a MRS.
The differences in power increases of brimmed DAWTs in a MRS
suggest the presence of a biased gap flow. However, neither the
velocity profile measured with hot wire nor a visualization using
smoke wire technique could confirm that a biased gap flow is
responsible for the differences in power increase.
Conventional wind turbines in a MRS also show differences in
power increase. Since biased gap flow is unlikely to occur with
conventional wind turbines, investigations need to be undertaken
to analyze the effect of the in-field blockage on the power increase
of each individual turbine in the MRS.
It can be concluded that brimmed DAWTs in MRS show dynamic
flow interactions. These can lead to an increase in power output
compared to a single-rotor DAWT. Future work includes analyzing
the in-field blockage and its effect on the power increase of each
wind turbine in a MRS with side-by-side and vertically spaced wind
turbines. Also, CFD simulations could provide more details on the
flow field, notably in regards to the gap between the turbines.
Acknowledgments
Acknowledgments go to Mr. Kimihiko Watanabe and Mr. Keiji
Matsushima for their contributions in the wind tunnel experi-
ments. Further, the Ministry of Education, Culture, Sports and Sci-
ence (MEXT) of Japan is acknowledged for the scholarship support
to the first author.
Abbreviations
A swept area of the rotor (m2)
AT turbine swept area (m2), for a conventional turbine A ¼ AT
AWT cross section area of the wind tunnel (m2)
b distance of the hot wire probe behind the brim of the
DAWT (m)
CP power coefficient
CP0i max. power coefficient of a single-rotor turbine
CPi max. power coefficient of a turbine in multi-rotor
configuration
CP0 average of max. power coefficients of all turbines in
single-rotor configuration
CP average of max. power coefficients of all turbines in
multi-rotor configuration
D brim diameter (m)
Drotor rotor diameter (m)
Dthroat throat diameter (m)
h brim height (m)
H projected width of the brimmed diffuser (m)
L characteristic length (m)
Lt length of the diffuser structure (m)
n number of turbines
PTurbine power of turbine (W)
PWind power in the wind (W)
Re Reynolds number
s gap width between turbines
s/D gap width between turbines s/brim diameter D
s/H gap width between turbines s/projected width H of the
brimmed diffuser
T/D space between center of flat plates T/width of plate D
U0 approaching wind speed (m/s)
UHW wind speed measured with hot wire (m/s)
DCP variation of power coefficient of MRS average
DCPi variation of power coefficient of individual turbines
l tip speed ratio
n kinematic viscosity (m2/s)
r air density (kg/m3)
References
[1] P. Jamieson, M. Branney, Multi-rotors; a solution to 20 MW and beyond?
Energy Procedia 24 (2012) 52e59, http://dx.doi.org/10.1016/
j.egypro.2012.06.086.
[2] G. Sieros, P. Chaviaropoulos, J.D. Sørensen, B.H. Bulder, P. Jamieson, Upscaling
wind turbines: theoretical and practical aspects and their impact on the cost
of energy, Wind Energy 15 (2012) 3e17, http://dx.doi.org/10.1002/we.527.
[3] M. Hofmann, I.B. Sperstad, Will 10 MW wind turbines bring down the oper-
ation and maintenance cost of offshore wind farms? Energy Procedia 53
(2014) 231e238, http://dx.doi.org/10.1016/j.egypro.2014.07.232.
[4] Danmarks Vindkrafthistoriske Samling (Danish Wind Historical Collection).
Available online: http://www.vindhistorie.dk/userfiles/downloads/faktablade/
Faktablad_2a.pdf (Accessed 31 May 2016).
[5] P. Jamieson, Multi rotor systems, in: first ed.Innovation in Wind Turbine
Design, vol. 2, John Wiley & Sons, Ltd, Chichester, UK, 2011, pp. 229e238,
http://dx.doi.org/10.1002/9781119975441.
[6] W. Heronemus, Pollution-free energy from offshore winds, in: 8th Annual
Conference and Exposition, Marine Technology Society, Washington, DC,
1972.
[7] P.T. Smulders, S. Orbons, C. Moes, Aerodynamic interaction of two rotors set
next to each other in one plane, in: European Wind Energy Conference, 1984,
pp. 529e533. Hamburg.
[8] D. Ransom, J.J. Moore, M. Heronemus-Plate, Performance of wind turbines in a
closely spaced array, Renew. Energy World 2 (2010) 32.
[9] P. Chasapogiannis, J.M. Prospathopoulos, S.G. Voutsinas, T.K. Chaviaropoulos,
Analysis of the aerodynamic performance of the multi-rotor concept, J. Phys.
Conf. Ser. 524 (2014) 012084, http://dx.doi.org/10.1088/1742-6596/524/1/
012084.
[10] S. McTavish, S. Rodrigue, D. Feszty, F. Nitzsche, An investigation of in-field
blockage effects in closely spaced lateral wind farm configurations, Wind
Energy 18 (2015) 1989e2011, http://dx.doi.org/10.1002/we.1806.
[11] A.R. Meyer Forsting, N. Troldborg, M. Gaunaa, The flow upstream of a row of
aligned wind turbine rotors and its effect on power production, Wind Energy
(2016), http://dx.doi.org/10.1002/we.1991.
[12] W.J. Rainbird, G.M. Lilley, A Preliminary Report on the Design and Perfor-
mance of a Ducted Windmill [Report 102], College of Aeronautics Cranfield,
1956.
[13] K. Foreman, B. Gilbert, R. Oman, Diffuser augmentation of wind turbines, Sol.
Energy 20 (1978) 305e311, http://dx.doi.org/10.1016/0038-092X(78)90122-
6.
[14] O. Igra, Research and development for shrouded wind turbines, Energy
Convers. Manag. 21 (1981) 13e48, http://dx.doi.org/10.1016/0196-8904(81)
90005-4.
[15] K. Abe, M. Nishida, A. Sakurai, Y. Ohya, H. Kihara, E. Wada, K. Sato, Experi-
mental and numerical investigations of flow fields behind a small wind tur-
bine with a flanged diffuser, J. Wind Eng. Industrial Aerodynamics 93 (2005)
951e970, http://dx.doi.org/10.1016/j.jweia.2005.09.003.
[16] M.O.L. Hansen, N.N. Sørensen, R.G.J. Flay, Effect of placing a diffuser around a
wind turbine, Wind Energy 3 (2000) 207e213, http://dx.doi.org/10.1002/
34 €ltenbott et al. / Renewable Energy 112 (2017) 25e34
U. Go
we.37.
[17] G.J.W. van Bussel, The science of making more torque from wind: diffuser
experiments and theory revisited, J. Phys. Conf. Ser. 75 (2007) 012010, http://
dx.doi.org/10.1088/1742-6596/75/1/012010.
[18] Y. Ohya, T. Karasudani, A shrouded wind turbine generating high output
power with wind-lens technology, Energies 3 (2010) 634e649, http://
dx.doi.org/10.3390/en3040634.
[19] S.H. Wang, S.H. Chen, The study of interference effect for cascaded diffuser
augmented wind turbines, in: The 7th Asia-Pacific Conf. On Wind Eng. Taipei,
Taiwan, 2009, pp. 8e12.
[20] Y. Ohya, J. Miyazaki, U. Go €ltenbott, T. Karasudani, S. Yoshida, Power
augmentation of shrouded wind turbines in a multi-rotor system. Proceedings
of 2nd International Conference on Next Generation Wind Energy (2nd
ICNGWE), Lund, Sweden, August 24-26, 2016.
[21] Boundary Layer Wind Tunnel, Wind Engineering Section, Research Institute
for Applied Mechanics, Kyushu University. Available online: http://www.riam.
kyushu-u.ac.jp/windeng/equipment.html (Accessed 06 July 2016).
[22] T. Matumiya, T. Kogaki, M. Iida, K. Kieda, A development of high performance
aerofoil, Turbomachinery 29 (2001) 519e524.
[23] Y. Ohya, Bluff body flow and vortex-its application to wind turbines, Fluid
Dyn. Res. 46 (2014) 061423, http://dx.doi.org/10.1088/0169-5983/46/6/
061423.
[24] Y. Ohya, A. Okajima, M. Hayashi, Wake interference and vortex shedding, in:
N.P. Cheremisinoff (Ed.), Encyclopedia of Fluid Mechanics vol. 8, Gulf Pub-
lishing Corporation, Houston, USA, 1989, pp. 323e389.
[25] Y. Nakamura, Y. Ohya, Vortex shedding from square prisms in smooth and
turbulent flows, J. Fluid Mech. 164 (1986) 77e89, http://dx.doi.org/10.1017/
S0022112086002471.