Accepted Manuscript

Experimental investigation of vertical-axis wind-turbine wakes in boundary layer flow

Vincent F-C. Rolin, Fernando Porté-Agel

PII: S0960-1481(17)31075-3
DOI: 10.1016/j.renene.2017.10.105
Reference: RENE 9393

To appear in: Renewable Energy

Received Date: 3 April 2017

Revised Date: 29 September 2017
Accepted Date: 29 October 2017

Please cite this article as: Rolin VF-C, Porté-Agel F, Experimental investigation of vertical-axis wind-
turbine wakes in boundary layer flow, Renewable Energy (2017), doi: 10.1016/j.renene.2017.10.105.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Experimental investigation of vertical-axis wind-turbine

wakes in boundary layer flow

PT
Vincent F-C. Rolin, Fernando Porté-Agela,∗
a Ecole Polytechnique Fédérale de Lausanne (EPFL), EPFL-ENAC-IIE-WIRE, Lausanne
1015, Switzerland

RI
SC
Abstract
In this experiment, a small scale vertical-axis wind-turbine (VAWT) is immersed
in a boundary-layer in a wind tunnel and stereo particle image velocimetry is

U
employed to quantify the 3D characteristics of the wake. The measurements
show that the wake is strongest behind the sector of the rotor which turns into
the wind. Two counter-rotating vortex pairs in the wake induce crosswind mo-
AN
tion which reintroduces streamwise momentum into the wake. Terms of the
mean kinetic energy budget are computed and demonstrate that this crosswind
flow has a significant influence on the redistribution of momentum in the wake.
A similar analysis of the turbulence kinetic energy budget identifies shearing
M
at the boundary of the wake as the primary contributor to the production of
turbulence. An analytical model is developed in order to obtain a theoreti-
cal basis from which to understand how the aerodynamic behavior of VAWTs
D

induces crosswind motion consistent with the production of counter-rotating

vortex pairs.
TE

Keywords: Atmospheric boundary layer, Vertical-axis wind-turbine,

Wind-turbine wakes, Stereo-PIV, Wind tunnel
EP

1. Introduction

Carbon emissions produced by the consumption of fossil fuels are a major

contributor to climate change, and dependence on oil combined with the volatil-
ity of oil prices is a growing source of geopolitical tension. As a result, renewable
C

5 energy has become an important field of study due to its demonstrated poten-
tial as a sustainable, emission-free alternative to fossil fuels. Wind power is one
AC

of the fastest growing and most cost-efficient of the different means of produc-
ing renewable energy available today. The largest horizontal-axis wind-turbines
(HAWTs) are able to generate 6MW of power and are able to extract close
10 to the theoretically maximum energy available from the wind, making them

∗ Corresponding author
Email address: fernando.porte-agel@epfl.ch (Fernando Porté-Agel)

Preprint submitted to Renewable Energy September 27, 2017

 ACCEPTED MANUSCRIPT

an appealing technology for large-scale energy production. Vertical-axis wind-

turbines (VAWTs), however, offer certain advantages over HAWTs which merit
study. They have, nonetheless, been the subject of significantly less research.

PT
One advantage of VAWTs is that the power they generate is independent of
15 the direction of the incoming wind as the turbine’s axis of rotation is oriented
perpendicular to the flow. This is an important advantage over HAWTs which
have to face the wind in order to attain peak performance. This requires so-

RI
phisticated technology in order to determine the direction of the incoming wind
and techniques to forecast how it will change in the future. An additional ben-
20 efit of VAWTs is that, due to their design, the generator can be placed below
the rotor which reduces loading on the tower relative to HAWTs. This config-

SC
uration presents a potential cost saving for the construction, installation, and
maintenance of the turbine.
The geometry of VAWTs may also prove advantageous to planning a wind-
25 farm. One way a rotor can capture wind energy is by presenting a larger frontal

U
area to the wind. HAWTs can enlarge this area by having a larger diameter,
but this means that the lateral distance between HAWTs must be increased to

30
AN
accommodate the larger rotor. VAWTs, by contrast, can expand the frontal
area they present to the wind by increasing the height of the rotor, rather than
the turbine diameter. This suggests that tall VAWTs can be placed closely to-
gether to conserve space and increase the power density of a wind farm. Dabiri
M
[1] demonstrated the potential advantage of a cluster closely spaced group of
counter-rotating VAWTs.
These advantages have made VAWTs the center of an expanding area of
35 study, yet a vast gulf of knowledge still exists between our understanding of
D

VAWT and HAWT wakes. Studies of HAWTs have shown that a thorough
characterization of the wake is critical to be able to plan a turbine array for op-
timal performance and longevity[2]. Specifically it is important to quantify the
TE

magnitude of turbulence, velocity deficit, and recovery rate of the wake behind
40 turbines upwind as it will influence those downwind. Understanding how the
wake is influenced by different inflow conditions such as a boundary layer rep-
resentative of atmospheric flow is important for scaling experiments to reality
EP

and for validating numerical studies. Many previous experimental investiga-

tions have been limited either by the implemented techniques which produced
45 measurements of the wake with poor spatial resolution [3–5]. Other studies
have been limited in scope, focused primarily on characterizing the complex dy-
C

namics of blade-blade interaction, vortex shedding, and the wake behavior only
within the rotor and near wake regions [6–13]. The far wake, however, is the
AC

most relevant zone regarding turbine-turbine interactions in wind-farms.

50 Of the few experimental studies which have investigated the far wake, most
have been limited to measurements over a single plane [4, 14] or did not fully
immerse the turbine in the type of boundary layer flow which would be encoun-
tered in a wind farm [15]. These studies have not reported consistent behavior of
the wake but tend to concur that the velocity deficit and turbulence in the wake
55 are not symmetrically distributed behind the turbine and that the structure and
magnitude of the wake is largely dependent on the rotor solidity and tip-speed

2
 ACCEPTED MANUSCRIPT

ratio. Khosravi et al. [16] found that the wake becomes more symmetric with
increasing tip-speed ratio.
Bergeles et al. [5] and Rolin and Porté-Agel [14] observed peak levels of

PT
60 velocity deficit behind the sector of the turbine’s rotation during which the
blades turn into the wind. The highest levels of turbulence in these studies
were in the shear layer observed between this region and the unperturbed flow.
Brochier et al. [4], Battisti et al. [9], and Posa et al. [17], however, observed

RI
the highest turbulence levels in the wake behind the opposite side of the rotor
65 where the blades turn with the wind. These observations attribute the high
turbulence in these regions to shear between the blades and surrounding flow
as well as the production of dynamic stall vortices.

SC
Some studies have observed phenomena which were not seen elsewhere. Ryan
et al. [15] reported the presence of a pair of counter-rotating vortices in the
70 wake of the VAWT but did not offer an explanation for their presence. Bas-
tankhah and Porté-Agel [18] observed a similar phenomenon in the wake of

U
yawed HAWTs and noted the similarity to the vortices associated with cross-
flow jets and concluded that the crosswind force exerted by the rotor induces

75
AN
the conditions required to produce these vortices.
Previous experiments have investigated the validity of small-scale empirical
studies on VAWTs. Measurements performed over a range of Reynolds numbers
by Bachant and Wosnik [19] indicate that the wake characteristics are largely
M
Reynolds insensitive. This suggests that empirical observations at low Reynolds
numbers are scalable. A comparison of measurements by Araya and Dabiri [20]
80 performed behind a rotor driven by the incoming wind, as opposed to a rotor
driven by a motor, shows that the wake observed in either case is indistinguish-
D

able so long as the tangential force produced by the interaction of the rotor and
flow is consistent with the turbine’s direction of rotation. Given this, the present
study focuses on characterizing the wake of a small-scale motorized VAWT in a
TE

85 boundary layer wind tunnel and how it develops downstream. The mechanisms
which govern the re-energization of the flow are identified and quantified via the
terms of the mean kinetic energy budget and momentum fluxes, and terms of
the turbulence kinetic energy budget equation are computed in order to iden-
EP

tify the mechanisms most responsible for the production of turbulence in the
90 wake. An analytical model is developed in order to develop a theoretical basis
to understand the lateral forces exerted on the flow believed to be the cause of
the counter-rotating vortex pair observed in the wake.
C

2. Experimental setup
AC

Measurements were taken using stereo particle image velocimetry (stereo-

95 PIV) behind a small scale H-type VAWT in the WIRE boundary layer wind
tunnel, depicted in Figure 1. The wind tunnel is powered by a 130 kW fan
which can generate winds up to 25 ms−1 . The test section is 28 m long with a
cross-section that is 2.57 m wide and 2.00 m tall. Measurements are conducted
20 m from the tunnel contraction such that a boundary layer develops naturally
100 over the floor of the facility. Vertical profiles of the mean velocity and streamwise

3
 ACCEPTED MANUSCRIPT

Honeycomb 5:1 Area Test section

and screens contraction

PT 2.71m
4.75m

2.71m

RI
4.00m

2.50m
SC
Diffuser Fan

U
Figure 1: Schematic of the boundary layer wind tunnel of the WIRE laboratory

are plotted in Figure 2. AN

turbulence intensity of the incoming flow, acquired via hot-wire anemometry,

The depth of the boundary layer at the rotor position is 40 cm and the veloc-
ity in the free stream is 11 ms−1 . The relationship between velocity and height
is roughly logarithmic up to height of 18 cm. From the log-law approximation,
M
U 1 z
= ln , (1)
u∗ κ z0
D

the values of the friction velocity and roughness are deduced as u∗ = 0.44 ms−1 ,
and z0 = 2.5 × 10−5 m respectively for κ = 0.4. This approximation is presented
in Figure 3.
TE

105

The rotor studied in this experiment is a three-bladed H-type Darrieus as

depicted in Figure 4. The blades are NACA0018 airfoils with a chord length of
3 cm and a span of 15.5 cm. The diameter of the rotor is 16.6 cm. The blockage
ratio of the rotor frontal area to the tunnel cross-section is 0.5 %. The turbine
EP

110 operates at a tip-speed ratio of λ = 1.1 based on a wind speed at the rotor
mid-span of 9.43 ms−1 . Two Reynolds numbers are used to characterize the
flow with respect to the turbine, based on the incoming wind speed and rotor
diameter, ReD = 100, 000, or on the airfoil chord-length and the speed relative
C

to the blades, ReC = 20, 000. The turbine is driven by a 12W brushless Maxon
115 EC45 flat motor in order to maintain a constant rotational velocity and level of
AC

performance. The motor was selected specifically for its low profile in order to
keep the rotor within the boundary layer and to avoid strong interactions with
the wind.
Stereo-PIV is implemented as illustrated in Figure 5. A dual pulsed laser
120 passes through a 10 mm cylindrical lens to illuminate a plane of interest across
the wind-tunnel. Two cameras are used to obtain stereoscopic images of the
plane, which is aligned normal to the streamwise direction. One camera is
placed upstream and the other downstream with respect to the measured flow,

4
 ACCEPTED MANUSCRIPT

PT
RI
SC
Figure 2: Profiles of the incoming boundary layer (left) velocity and (right) turbulence inten-

U
sity. The rotor is illustrated to scale for reference.

1.1
AN
1
M
U=UH

0.9
0.8
D

0.7
TE

100
Z=D

Figure 3: The measured (dots) and logarithmic fit (solid line) of velocity in the boundary
EP

layer.
C
AC

Figure 4: Model H-type vertical-axis wind-turbine

5
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Figure 5: Top-down view of stereo-PIV configuration.

with an angle of roughly 87 ◦ between them and a distance of approximately

M
125 1.7 m from the center of the plane. This configuration allows the major flow
features to be captured over a 3D × 4D area. Post-processing is performed over
32 × 32 pixel interrogation windows with 50 % overlap, resulting in 250 × 168
D

points of data with a resolution of 0.02 D. The sampling frequency is roughly

10 Hz, with 200µs between exposures. Measurements are taken of the incoming
flow and of the wake at 2, 3, 4, 6, 8, and 10 rotor diameters from the turbine.
TE

130

3. Results

In this section the data acquired by stereo-PIV are analyzed and reported.
EP

Mean quantities are computed as time averages of snapshots of the flow. The
shape and structure of the wake is presented in Figure 6, which shows mean
135 velocity deficit over all measured planes.
The velocity deficit at each individual measurement plane is presented in
C

Figure 7. Comparing the velocity deficit to the position of the rotor indicates
that the far wake is deflected towards the negative-y direction, and recovers more
AC

momentum above the rotor midspan than below. Figures 8 presents a horizontal
140 vertical profile of the velocity deficit, showing clearly that the velocity deficit in
the wake unevenly distributed such that the momentum behind the negative-y
sector of the rotor which turns into the wind is significantly lower than behind
the positive-y sector. The vertical profile of velocity deficit in Figure 9 shows
that the higher region of the wake returns to the inlet condition more quickly
145 than the lower portion.

6
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Figure 6: Contours of streamwise velocity deficit normalized by the incoming wind speed at
the midspan over each measurement plane, with the rotor situated at the origin. The turbine’s
direction of rotation is indicated by a black arrow
M

1 ! U =UH
D

0.00 0.25 0.50 0.75 1.00

TE

x/D = 2 x/D = 3 x/D = 4

2
z=D

1
EP

0
x/D = 6 x/D = 8 x/D = 10
2
z=D

1
AC

0
-1 0 1 -1 0 1 -1 0 1
y=D y=D y=D

Figure 7: Contours of streamwise velocity deficit at different downstream positions normalized

by the incoming wind speed at the midspan. In-plane velocity vectors are overlaid. The outline
of the rotor is delineated by a black box.

7
 ACCEPTED MANUSCRIPT

1 x/D = 2 x/D = 3 x/D = 4

PT
y=D

RI
-1
1 x/D = 6 x/D = 8 x/D = 10

SC
y=D

-1

U
0 0.5 1 0 0.5 1 0 0.5 1
1 ! U =UH 1 ! U =UH 1 ! U =UH

AN
Figure 8: Crosswind profiles of the streamwise momentum deficit normalized by the incoming
velocity at the midspan. Dotted lines denote the lateral edges of the VAWT rotor for reference
M
D

x/D = 2 x/D = 3 x/D = 4

2
TE
z=D

0
EP

x/D = 6 x/D = 8 x/D = 10

2
z=D

1
C

0
AC

0 1 2 0 1 2 0 1 2
U =UH U =UH U =UH

Figure 9: Vertical profiles of streamwise velocity normalized by the incoming wind-speed at

the midspan. The dashed line represents the incoming boundary layer flow. Dotted lines
denote the upper and lower edges of the VAWT rotor for reference

8
 ACCEPTED MANUSCRIPT

!x # DUH!1

PT
-1.00 -0.75 -0.50 -0.25 0.00 0.25 0.50 0.75 1.00
x/D = 2 x/D = 3 x/D = 4
2
z=D

RI
0
x/D = 6 x/D = 8 x/D = 10
2

SC
z=D

U
-1 0 1 -1 0 1 -1 0 1
y=D y=D y=D

AN
Figure 10: Contours of vorticity and in-plane velocity vectors at different downstream positions
normalized by the incoming wind speed at the midspan and rotor diameter. In-plane velocity
vectors are overlaid. The outline of the rotor is delineated by a black box.
M
The vorticity at each measurement plane is presented in Figure 10 with the
in-plane velocity vectors overlaid. This figure clearly indicates the presence of
two pairs of counter-rotating vortices. The pairs are of different magnitudes,
D

with those in the wake of the negative-y sector of the rotor which turns into
150 the wind being significantly larger and stronger than those in the wake of the
positive-y sector. These vortices appear to induce significant crosswind motion.
TE

The added turbulence kinetic energy, or TKE, is defined as the difference

between the TKE in the wake and the incoming boundary layer, T KE+ =
T KEw − T KE0 . Figure 11 shows the magnitude of added TKE at each mea-
155 surement plane. It is clear that the highest levels of TKE occur at the periphery
EP

of the wake where there is strong shear. Figure 12 shows profiles of T KE+
which show that the stronger velocity gradient on the negative-y boundary of
the wake produces significantly stronger turbulence. The vertical profile of TKE
in the wake, shown in Figure 13, shows that the faster moving flow higher in
C

160 the boundary layer causes a stronger velocity gradient at the upper boundary
of the wake, leading to stronger turbulence.
AC

The high velocity deficit close to the ground, and high turbulence level near
the top of the wake is consistent with the previously observed behavior of the
wake behind a HAWT in boundary layer flow[21]. The following sections will
165 analyze the mechanisms underlying the recovery of momentum and production
of TKE in the wake.

9
 ACCEPTED MANUSCRIPT

T KE+ # UH!2

PT
x10-2
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00
x/D = 2 x/D = 3 x/D = 4
2

RI
z=D

SC
0
x/D = 6 x/D = 8 x/D = 10
2
z=D

U
0
-1 0
y=D
1
AN -1 0
y=D
1 -1 0
y=D
1

Figure 11: Contours of T KE+ at different downstream positions normalized by the incoming
wind speed at the midspan. In-plane velocity vectors are overlaid. The outline of the rotor is
M
delineated by a black box.
D

1 x/D = 2 x/D = 3 x/D = 4

TE
y=D

0
EP

-1
1 x/D = 6 x/D = 8 x/D = 10
y=D

0
AC

-1
0 0.025 0.05 0 0.025 0.05 0 0.025 0.05
T KE+ # UH!2 T KE+ # UH!2 T KE+ # UH!2

Figure 12: Crosswind profiles of TKE in the wake normalized by the incoming velocity at the
midspan. Dotted lines denote the lateral edges of the VAWT rotor for reference

10
 ACCEPTED MANUSCRIPT

x/D = 2 x/D = 3 x/D = 4

z=D 2

PT
1

0
x/D = 6 x/D = 8 x/D = 10

RI
2
z=D

SC
0
0 0.025 0.05 0 0.025 0.05 0 0.025 0.05
+T KE # UH!2 +T KE # UH!2 +T KE # UH!2

U
Figure 13: Vertical profiles of TKE in the wake normalized by the incoming wind-speed at the
midspan. The dashed line represents the TKE in the incoming boundary layer flow. Dotted
AN
lines denote the upper and lower edges of the VAWT rotor for reference

3.1. Mean kinetic energy budget

The budget of mean kinetic energy, or MKE, is a useful tool for identifying
M
phenomena responsible for the production and destruction of kinetic energy.
The budget equation decomposes the conservation of MKE into components
which each describe different mechanisms responsible for re-energizing the wake
D

and redistributing momentum. Expressed using Einstein index notation, the

equation for the mean kinetic energy budget is
2 2
TE

1 ∂U i 1 ∂U i U i ∂P
+ Uj = − gδi3 U i −
2 ∂t 2 ∂xj ρ ∂xj
| {z } | {z } | {z } | {z }
I II III IV
(2)
2
∂ Ui ∂U i ∂u0i u0j U i
EP

+ νU i 2 + u0i u0j − .
∂xj ∂xj ∂xj
| {z } | {z } | {z }
V VI V II

Given statistically stationary conditions, term I in Equation 2 can be ne-

C

glected, term III is ignored following the Boussinesq approximation as the flow
170 is relatively hydro-statically invariant, and term V regarding molecular dissipa-
AC

tion is omitted from this discussion as it was found to contribute little to the
production of MKE. Term IV could not be computed due to the lack of pres-
sure data collected in the present experiment. The streamwise gradients are
not calculated due to the large distances between measurement planes, though
175 these gradients tend to have a limited influence on the production of TKE in
comparison to the in-plane terms.
Term II represents the quantity of mean kinetic energy advected by the mean
flow, shown in Figure 14. There is a region of positive advection in the upper

11
 ACCEPTED MANUSCRIPT

2
!1=2 U j @U i =@xj # DUH!3
x10-1

PT
-2.00 -1.00 0.00 1.00 2.00
x/D = 2 x/D = 3 x/D = 4
2
z=D

RI
0
x/D = 6 x/D = 8 x/D = 10
2

SC
z=D

U
-1 0 1 -1 0 1 -1 0 1
y=D y=D y=D

AN
Figure 14: Advection of MKE at different downstream positions normalized by the rotor
diameter and incoming flow at the midspan height. In-plane velocity vectors are overlaid.
The outline of the rotor is delineated by a black box..
M
region of the wake, where the combined action of the vortices induces the highly
180 energetic unperturbed flow above to moved down into the wake. The advection
of MKE is negative at the interface of the unperturbed flow and the wake behind
D

the negative-y sector of the rotor as the wake is deflected into the unperturbed
flow. The kinematic advective flux of streamwise momentum, shown in Figure
15, illustrates the impact of the counter-rotating vortex pair on transporting
TE

185 momentum into the wake.

Turbulence also tends to play a large role in mixing momentum and re-
energizing the wake. The total turbulence flux of streamwise momentum, shown
in Figure 16, demonstrates that the high shear at the periphery of the wake
EP

leads to turbulence-induced mixing of momentum. Comparing the magnitude

190 of momentum entrained via turbulence and the mean flow shows that the mean
flow plays the more significant role in reintroducing momentum into the wake.
Term VI, shown in Figure 17, reflects the interaction of turbulence and the
C

mean flow. This term is predominantly negative, and overlaps with the regions
of high TKE, indicating that energy mechanically produced as turbulence where
AC

195 the velocity gradient acts against the Reynolds stresses is lost from the mean
flow. Term VII, shown in Figure 18, illustrates another mechanism by which the
interaction of turbulence and the mean flow produces MKE. This term achieves
peak values where the gradient of the turbulence flux of momentum is highest
and is positive where the velocity decreases at a decreasing rate.
200 The MKE budget analysis reveals two phenomena that play important roles
in re-energizing the wake. The mean flow induced by the counter rotating vortex
pairs transports momentum into the wake from the faster moving flow and

12
 ACCEPTED MANUSCRIPT

1! "2 ! "2 2
UV + UW # UH!4
x10-2

PT
0.00 0.38 0.75 1.13 1.50 1.88 2.25 2.63 3.00
x/D = 2 x/D = 3 x/D = 4
2

RI
z=D

SC
x/D = 6 x/D = 8 x/D = 10
2
z=D

U
0
-1 0 1 -1 0 1 -1 0 1
y=D
AN y=D

Figure 15: Advective flux of streamwise momentum normalized by the incoming wind speed
y=D

at the midspan height. In-plane velocity vectors are overlaid. The outline of the rotor is
delineated by a black box..
M

1 2
2 2
D

u0 v 0 + u0 w 0 # UH!4
x10-4
TE

0.00 0.88 1.75 2.62 3.50 4.38 5.25 6.12 7.00

x/D = 2 x/D = 3 x/D = 4
2
z=D

1
EP

0
x/D = 6 x/D = 8 x/D = 10
2
C
z=D

1
AC

0
-1 0 1 -1 0 1 -1 0 1
y=D y=D y=D

Figure 16: Turbulence flux of streamwise momentum normalized by the incoming wind speed
at the midspan height. In-plane velocity vectors are overlaid. The outline of the rotor is
delineated by a black box..

13
 ACCEPTED MANUSCRIPT

u0i u0j @Ui =@xj # DUH!3

x10-1

PT
-1.00 -0.50 0.00 0.50 1.00
x/D = 2 x/D = 3 x/D = 4
2

RI
z=D

SC
x/D = 6 x/D = 8 x/D = 10
2
z=D

U
0
-1 0 1 -1 0 1 -1 0 1
y=D
AN y=D y=D

Figure 17: Production of MKE via turbulence at different downstream normalized by the
rotor diameter and incoming flow at the midspan height with contours of velocity deficit and
in-plane velocity vectors overlaid. The outline of the rotor is delineated by a black box.
M
D

! @u0i u0j Ui =@xj # DUH!3

x10-1
TE

-2.00 -1.00 0.00 1.00 2.00

x/D = 2 x/D = 3 x/D = 4
2
z=D

1
EP

0
x/D = 6 x/D = 8 x/D = 10
2
C
z=D

1
AC

0
-1 0 1 -1 0 1 -1 0 1
y=D y=D y=D

Figure 18: Production of MKE via turbulence at different downstream normalized by the rotor
diameter and incoming flow at the midspan height. In-plane velocity vectors are overlaid. The
outline of the rotor is delineated by a black box..

14
 ACCEPTED MANUSCRIPT

advects wake into the unperturbed flow. Turbulence plays a less straightforward
role in the MKE budget. The mechanical energy produced as turbulence is lost
205 from the mean flow where there is strong shear. The turbulent fluctuations

PT
on the edges of the wake also cause momentum to be entrained into the wake.
Comparing the mean and turbulence fluxes of momentum demonstrates that
the mean flow plays the larger role.

RI
3.2. Turbulence kinetic energy budget
As previously discussed the highest levels of TKE observed in the wake of the
VAWT coincide with the strongest velocity gradients on the boundaries of the

SC
wake. In order to identify the mechanism responsible for producing turbulence
we turn to the TKE budget equation, given below,

∂e ∂e g  0 0 ∂Ui
+ Uj = + δi3 ui θv − u0i u0j
∂t ∂xj θ ∂xj

U
|{z} | {z } | v {z } | {z }
I II III IV
  (3)
−AN
∂ uj e
∂xj
0
−
1 u0i p0
ρ ∂xj
| {z } | {z } |{z}
− ε ,

V VI V II
M
210 where e represents TKE.
Some terms of Equation 3 can be neglected given certain assumptions re-
garding the flow. Terms I and III of the budget equation yield negligible contri-
butions to the production of TKE as the wake is considered both statistically
D

stationary and isothermal. Term VI which is related to pressure oscillations

215 could not be calculated due to the lack of pressure measurements, and the rate
TE

of turbulence dissipation represented in Term VII could not be quantified due

a similar lack of relevant data. The contribution of turbulent transport of TKE
is found to be negligible in comparison with that of the other terms, and is
omitted from this section. As mentioned earlier streamwise gradients are not
EP

220 presented due to the large distances between measurement planes.

The remaining terms include the in-plane contributions to TKE production
due to advection, and mechanical production (terms II and IV, respectively.)
One should note that the production term is actually the opposite of term VI
C

in Equation 2, such that the shearing which detracted form the MKE budget
225 contributes to the production of TKE. The last remaining term to consider is
that related to the advection of turbulence, the magnitude of which is presented
AC

in Figure 19. By comparison, it is clear that TKE is predominantly generated

via shear between the wake and the unperturbed flow.
The amount of turbulence transported across the wake by the motion in-
230 duced by the counter-rotating vortex pairs is of secondary importance. This
behavior is consistent with the production of turbulence in shear layers such as
those in the wakes of HAWTs.

15
 ACCEPTED MANUSCRIPT

!U j @e=@xj # DUH!3
x10-2

PT
-3.00 -1.50 0.00 1.50 3.00
x/D = 2 x/D = 3 x/D = 4
2
z=D

RI
1

SC
x/D = 6 x/D = 8 x/D = 10
2
z=D

U
0
-1 0 1 -1 0 1 -1 0 1
y=D AN y=D y=D

Figure 19: Advection of TKE at different downstream positions normalized by the rotor
diameter and incoming flow at the midspan height. In-plane velocity vectors are overlaid.
M
The outline of the rotor is delineated by a black box..

3.3. Counter-rotating vortex pairs

D

As stated in the previous section, the wake of the VAWT features two pairs
235 of counter rotating vortices, clearly visible in Figure 10. Via the analysis of
the MKE budget, these vortices are shown to be the principal cause of wake
TE

re-energization. It is therefore critically important to understand the cause of

these vortices and under what circumstances will their effect be augmented or
diminished. Ryan et al. [15] observed a similar CVP in the wake of a VAWT at
low Reynolds numbers using magnetic resonance velocimetry and identify them
EP

240

as tip vortices caused by the lift-producing pressure differential over the rotor
blades. Hofemann et al. [7] followed the trajectory of rotor tip vortices using
PIV techniques in the rotor near-wake, and Tescione et al. [12] observed their
presence and influence on the flow up to a distance of three diameters from the
C

245 turbine, at which point these vortices tend to roll-up and dissipate. Studies
of HAWTs by Chamorro and Porté-Agel [22] demonstrate that boundary layer
AC

turbulence causes rotor-tip vortices to break down earlier than in smooth flow.
Given this information and the fact that previously observed rotor tip vortices
are much smaller than the CVP in the present experiment and those reported
250 by Ryan et al. [15], it is unlikely that these counter-rotating pairs are in fact tip
vortices.
Bastankhah and Porté-Agel [18] observed a similar CVP in the wake of
yawed HAWTs in wind-tunnel experiments. They noted that this phenomenon
is similar to the CVP associated with crossflow jets, a scenario in which a jet

16
 ACCEPTED MANUSCRIPT

255 is oriented orthogonally to a uniform flow. A literature review of crossflow

jets can be found in Mahesh [23], though it must be noted that the literature
offers many possible reasons for why crossflow gives rise to a CVP without

PT
proposing a definitive explanation. Bastankhah and Porté-Agel posited that
counter-rotating vortex pairs occur in the presence of a varying distribution of
260 lateral velocity as a necessary result of continuity. A more complete explanation
of this mechanism can be found in the original work. Briefly, in the case of

RI
HAWTs, introducing a yaw angle results in a crosswind force being exerted on
the flow which in turn induces lateral velocity by conservation of momentum.
Due to continuity, the crosswind gradient of lateral velocity induces a vertical
gradient of vertical velocity; these gradients describe the crosswind vortices

SC
265

observed in the flow. Due to the rotation of the VAWT, the blades exert both
a streamwise and a lateral force on the wind. As with yawed HAWTs, this
crosswind force introduces crosswind momentum to the flow, and consequently
gives rise to the circumstances we have associated with the production of a CVP.

U
270 3.4. Analytical model
AN
In order to investigate the underlying aerodynamics which cause the VAWT
rotor to exert a lateral force on the flow, an analytical model is developed
based on a blade-element-momentum techniques adapted for VAWTs and yawed
actuators. The Double-Multiple Streamtube Model, described by Paraschivoiu
M
275 [24], divides the VAWT rotor into a series of parallel streamtubes aligned with
the incoming flow in order to more accurately capture the changing aerodynamic
behavior of the blades as they turn. Each streamtube features two actuators
for the two regions where the blades’ trajectory crosses the streamtube. In this
D

way the downwind sector of the rotor will interact with the wake of the upwind
280 sector. Given the velocity relative the blades, the values of lift and drag can be
determined based on tabulated data found in the work of Sheldahl and Klimas
TE

[25] and then modified to account for dynamic stall using the MIT model, as
implemented by Shamsoddin and Porté-Agel [26].
The flow relative to a blade in motion and the resulting forces are illustrated
in figure 20. It is clear from the figure that the force on the blade, F~ , is
EP

285

not necessarily aligned with the incoming flow, U ~ ∞ . The model is modified in
order to incorporate the crosswind component of the lift and drag forces which
previously were neglected. This is achieved by yawing the actuators in each
streamtube by an angle γ = arctan F~ · y/F~ · x. A diagram of the model is
C

290 provided in figure 21 The actuator induces a velocity aU∞ in the same direction
as the force vector, where a is an induction factor. The value of the induction
AC

factor is related to the thrust coefficient of the actuator by CT = 4a (cos γ − a),

and the thrust coefficient CT is equivalent to the total aerodynamic force exerted
by an actuator normalized by the incoming wind speed and the area of the
295 streamtube. The calculation is iterated, incorporating the induced flow into the
velocity relative to the turbine blade to recalculate the lift and drag. A more
detailed explanation of the model can be found in the Appendix.
By including the lateral component of the aerodynamic forces a prediction
can be made regarding the direction and magnitude of the crosswind momentum

17
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Figure 20: The flow relative to the blade and the resulting forces.
M
D
TE
C EP
AC

Figure 21: Schematic illustration of the Double Multiple Streamtube model with yawed actu-
ators.

18
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
Figure 22: Overhead view of the 2D VAWT rotor analyzed using the yaw-adapted Double-
Multiple Streamtube model with the direction and magnitude of the velocity induced by the
rotor illustrated by the blue vectors. The direction and magnitude of the incoming flow is
denoted by the red arrow. The VAWT’s direction of rotation is illustrated by the circular
D

black arrow.
TE

300 introduced into the flow. This 2D analysis of the VAWT produces the distribu-
tion of velocity induced by the rotor shown in Figure 22. The induced velocity
predicted by the model is consistent with the conditions which produce CVPs
based on the previously presented continuity argument. This is an important
EP

result, as it demonstrates that VAWTs induce forces on the flow in a manner

305 akin to yawed HAWTs. The two turbines therefore produce wakes with similar
characteristics, such as CVP and the deflection of the momentum deficit from
the centerline. It remains to be seen how different VAWT design parameters
C

and operating conditions will alter the similarity between VAWTs and yawed
HAWTs.
AC

310 4. Summary

Using stereo-PIV techniques in a wind tunnel, the wake behind a small

VAWT immersed in a boundary layer is quantified and analyzed. Several in-
teresting characteristics of the wake are observed. Possibly the most notable
feature of the flow is the presence of two pairs of counter-rotating vortices at
315 the edges of the wake. Analyzing the MKE budget and fluxes of momentum

19
 ACCEPTED MANUSCRIPT

in the wake indicates that these CVPs are the most important mechanism in
introducing momentum to the wake. These counter-rotating vortex pairs tend
to occur where there is a variation in crossflow momentum and have also been

PT
observed in the wakes of yawed HAWTs due to the lateral force these rotors pro-
320 duce. An analytical model is developed based on the Double-Multiple Stream-
tube model commonly applied to VAWTs and is adapted to incorporate yawed
actuators. The model demonstrates that VAWTs are expected to exert lateral

RI
forces and induce flow consistent with the production of CVPs observed in the
wake. VAWT wakes appear to have much in common with the wakes of yawed
325 HAWTs, including not only the production of CVPs but also the lateral de-
flection of the wake center in the same direction as the velocity induced by the

SC
rotor.
Due to the boundary layer flow, the wake recovers asymmetrically as more
momentum is entrained from the faster moving flow above the wake by the
330 counter-rotating vortices. The high velocity gradient on the boundary of the

U
wake behind the negative-y sector of the rotor gives rise to peak turbulence
levels and encourages mixing in this region. Through the previously discussed

335
AN
analysis of the MKE budget, it appears that the turbulence flux of streamwise
momentum does not play as important a role as the mean flux in terms of
re-energizing the wake. The boundary layer also causes the turbulence added
to the flow to be greater at the upper boundary between the wake and the
M
unperturbed flow than at the lower boundary due to the higher shear in this
region as a result of the faster moving flow higher in the boundary layer.
It remains unclear why there is such a discrepancy between VAWT wakes
340 observed in literature. In the case studied here it seems clear the turbulence
D

is primarily produced via shear, and that the stronger the velocity deficit, the
higher the turbulence. It may prove to be the case that the low tip-speed ratio
implemented in this study results in highly asymmetric aerodynamic behavior,
TE

as the flow relative to the blade will fluctuate highly as the VAWT rotor spins.
345 The analytical model developed here, once validated, may prove useful to pre-
dicting how tip-speed ratio is related to these oscillatory aerodynamics and wake
asymmetry.
EP

Acknowledgments:. This research was supported by the Swiss National Science

Foundation [grant 206021 144976], the Swiss Federal Office of Energy [grant
350 SI/501337-01], and the Swiss Innovation and Technology Committee (CTI)
C

within the context of the Swiss Competence Center for Energy Research ’FU-
RIES: Future Swiss Electrical Infrastructure’.
AC

Appendix A.

We examine the aerodynamics of vertical-axis wind-turbines in order to in-

355 vestigate the lateral forces produced by the rotor. To this purpose we ap-
ply a modified version of the Double-Multiple Streamtube Model described by
Paraschivoiu [24]. This model segments the rotor into a series of parallel stream-
tubes aligned with the incoming flow. Each streamtube features two actuators,

20
 ACCEPTED MANUSCRIPT

which represent the two zones where the blades intersect the flow. The present
360 study suggests modifying the actuators to account for the influence of lateral
forces.

PT
The incoming velocity is defined as
~ = U∞ î,
U (A.1)

where U∞ is the magnitude of the incoming flow, and î is the unit vector aligned

RI
with the streamwise flow. We consider an actuator which crosses the streamtube
with a yaw angle γ. During the upwind portion of the cycle, where the incoming
365 flow is unperturbed, we assume an induction factor a which induces a velocity

SC
normal to the actuator,
 
~ a = −aU∞ cos γ î + sin γ ĵ
U (A.2)

We assign a coordinate system to the blade’s frame of reference such that the

U
unit vector aligned with the chord line is

AN
ĉ = sin θî − cos θĵ, (A.3)

where θ is the azimuthal angle between the blade’s position and the abscissa,
and ĵ is the unit vector perpendicular to the incoming flow in the absolute frame
M
of reference. The unit vector normal to the chord is therefore

n̂ = cos θî + sin θĵ. (A.4)

D

The tangential velocity of a blade is, by definition, λU∞ , where λ is the tip-
speed-ratio, or the ratio of the blade tangential velocity tot the incoming flow.
The total relative velocity with respect to an individual blade is therefore
TE

~ =U
W ~ +U
~ a + λU∞ ĉ. (A.5)

The angle of attack between the flow relative to the blade and the blade’s
EP

direction of motion is !
W~ · n̂
α = arctan . (A.6)
~ · ĉ
W
The coefficients of lift and drag, Cl and Cd respectively, are extrapolated from
C

data provided for NACA0018 airfoils by Kumar et al. [27] and Sheldahl and
Klimas [25]. Dynamic stall effects are reproduced via the modified MIT model
AC

developed by Noll and Ham [28] which is a modification of the original MIT
model[29]. Shamsoddin and Porté-Agel [30] previously implemented this model
in an LES study of VAWT wakes and provided a thorough explanation of how
to do so for others to follow. The drag force is aligned with the flow relative to
the blade and therefore has a unit vector
~
W
dˆ = . (A.7)
||W ||

21
 ACCEPTED MANUSCRIPT

The unit vector of lift is normal to the drag vector,

~
ˆl = W⊥ . (A.8)

PT
||W ||

The total force on the blades is

1  
T~ = ρcHB||W ||2 Cl ˆl + Cd dˆ ,

RI
(A.9)
2
where B, c, H are respectively the number of blades, the blade chord length,
and the blade span. Reciprocally, the thrust force exerted on the streamtube

SC
is −T~ . This vector consists of both the streamwise and lateral components of
thrust. The yaw angle then follows as the argument of the thrust vector
!
−T~ · ĵ
γ = arctan . (A.10)

U
−T~ · î

AN
For a streamtube encapsulating the sector of the rotor between θ − ∆θ and
θ + ∆θ, the overall actuator thrust is the mean thrust over the streamtube as a
portion of the number of streamtubes
Z θ+∆θ
1 1
q
M
T = ρcHB ||W ||2 Cl2 + Cd2 dθ. (A.11)
2 2N ∆θ θ−∆θ

By conservation of momentum, the trust exerted by the actuator should be

D

equivalent to the product of the velocity normal to the actuator and the total
change in mass-flow-rate in the streamtube,
TE

T = ρAs (U∞ cos γ − aU∞ ) (U∞ − (1 − 2a) U∞ ) , (A.12)

where As is the streamtube area. For streamtubes aligned with the x-axis,
the upwind azimuthal angle ranges between π2 and 3π
2 , the frontal area of a
streamtube centered around azimuthal angle θ is
EP

θ ∆θ
As (θ) = DH| cos sin |, (A.13)
2 2
C

where D is the rotor diameter. Equating this expression to the magnitude of

the thrust in Equation A.11 and normalizing by 12 ρAs U∞ 2
yields the following
relationship for the induction factor and thrust for a single streamtube
AC

R θ+∆θ  ||W || 2 p
σ θ−∆θ U∞ Cl2 + Cd2 dθ
CT = 4a (cos γ − a) = , (A.14)
2N ∆θ| cos θ2 sin ∆θ
2 |

Bc
where σ is the rotor solidity, defined as D . For values of thrust CT > 0.96,
Glauert’s empirical correction is applied.

22
 ACCEPTED MANUSCRIPT

Equations A.2 - A.14 can be iterated until convergence. This process is

repeated for the downwind portion of the rotor, using
h i
~ + 2aU∞ − cos γ î − sin γ ĵ

PT
U (A.15)

as the incoming streamtube velocity. The magnitude of this incoming velocity

370 should also be used when normalizing values with respect to the downwind

RI
actuators. This procedure can be used to predict the direction of the velocity
induced by the VAWT.
This model has some shortcomings and limitations such as the fact that it
does not account for the deflection or expansion of the streamtubes nor the

SC
375 influence of one streamtube on another. Nonetheless, due to the fluctuating
orientation of the blades with the incoming wind, relatively large lateral forces
can be incurred and cannot be neglected as they will influence the locally in-
duced velocity. This model takes steps to incorporate this lateral force into the

U
momentum balance for the VAWT and improve its robustness as a means of
380 evaluating the performance of a turbine.

References
AN
[1] J. O. Dabiri, Potential order-of-magnitude enhancement of wind farm
M
power density via counter-rotating vertical-axis wind turbine arrays, Jour-
nal of Renewable and Sustainable Energy 3 (4).
385 [2] R. B. Cal, J. Lebrón, L. Castillo, H. S. Kang, C. Meneveau, Experimental
D

study of the horizontally averaged flow structure in a model wind-turbine

array boundary layer, Journal of Renewable and Sustainable Energy 2
(2010) 1–25, ISSN 19417012.
TE

[3] R. J. Muraca, R. J. Guillotte, Wind tunnel investigation of a 14ft verit-

390 cal axis windmill, Tech. Rep., NASA Langley Research Center, Hampton,
Virginia, 1976.
EP

[4] G. Brochier, P. Fraunie, C. Beguier, I. Paraschivoiu, Water Channel Ex-

periments of Dynamic Stall on Darrieus Wind Turbine Blades, Journal of
Propulsion 2 (5) (1986) 445–449.
C

395 [5] G. Bergeles, A. Michos, N. Athanassiadis, Velocity vector and turbulence

in the symmetry plane of a Darrieus wind generator, Journal of Wind
AC

Engineering and Industrial Aerodynamics 37 (1) (1991) 87–101.

[6] N. Fujisawa, S. Shibuya, Observations of dynamic stall on Darrieus wind
turbine blades, Journal of Wind Engineering and Industrial Aerodynamics
400 89 (2) (2001) 201–214.

[7] C. Hofemann, C. J. S. Ferreira, K. R. Dixon, G. J. W. van Bussel, G. a. M.

van Kuik, A. Maria, F. Scarano, G. V. Bussel, G. V. Kuik, 3D Stereo

23
 ACCEPTED MANUSCRIPT

PIV Study of Tip Vortex Evolution on a VAWT, in: Wind Energy Confer-
ence and Exhibition EWEC, European Wind Energy Association EWEA,
405 Brussels, 1–8, 2008.

PT
[8] C. J. S. Ferreira, The near wake of the VAWT 2D and 3D views of the
VAWT aerodynamics, Ph.D. thesis, Universidade Tecnica de Lisboa, 2009.

[9] L. Battisti, L. Zanne, S. Dell’Anna, V. Dossena, G. Persico, B. Paradiso,

RI
Aerodynamic Measurements on a Vertical Axis Wind Turbine in a Large
410 Scale Wind Tunnel, Journal of Energy Resources Technology 133 (3) (2011)
031201.

SC
[10] J. Bossard, Caracterisation experimentale du decrochage dynamique dans
les hydroliennes a flux transverse par la methode PIV (Particle Image
Velocimetry). Comparaison avec les resultats issus des simulations nu-
415 meriques, Ph.D. thesis, Universite de Grenoble, 2013.

U
[11] L. Beaudet, Etude expérimentale et numérique du décrochage dynamique
sur une éolienne à axe vertical de forte solidité, Ph.D. thesis, Université de
Poitiers, 2014. AN
[12] G. Tescione, D. Ragni, C. He, C. J. S. Ferreira, G. J. W. van Bussel, Near
420 wake flow analysis of a vertical axis wind turbine by stereoscopic particle
M
image velocimetry, Renewable Energy 70 (2014) 47–61.

[13] P. Bachant, M. Wosnik, Characterising the near-wake of a cross-flow tur-

bine, Journal of Turbulence 16 (4) (2015) 392–410.
D

[14] V. Rolin, F. Porté-Agel, Wind-tunnel study of the wake behind a vertical

425 axis wind turbine in a boundary layer flow using stereoscopic particle im-
TE

age velocimetry, Journal of Physics: Conference Series 625 (2015) 012012,

ISSN 1742-6588, URL http://stacks.iop.org/1742-6596/625/i=1/a=
012012?key=crossref.276a4e4067e30f10c25e95f0c50e7beb.
EP

[15] K. J. Ryan, F. Coletti, C. J. Elkins, J. O. Dabiri, J. K. Eaton, Three-

430 dimensional flow field around and downstream of a subscale model rotating
vertical axis wind turbine, Experiments in Fluids 57 (3) (2016) 1–15.

[16] M. Khosravi, P. Sarkar, H. Hu, An Experimental Investigation on the Wake

C

Characteristics of a Wind Turbine in an Atmospheric Boundary Layer

Wind, AIAA AIAA 2016 (1732), ISSN 08899746.
AC

435 [17] A. Posa, C. M. Parker, M. C. Leftwich, E. Balaras, Wake structure of a

single vertical axis wind turbine, International Journal of Heat and Fluid
Flow 61 (Part A) (2016) 75–84.
[18] M. Bastankhah, F. Porté-Agel, Experimental and theoretical study of wind
turbine wakes in yawed conditions, Journal of Fluid Mechanics .

24
 ACCEPTED MANUSCRIPT

440 [19] P. Bachant, M. Wosnik, Effects of Reynolds number on the energy con-
version and near-wake dynamics of a high solidity vertical-axis cross-flow
turbine, Energies 9 (2) (2016) 1–18.

PT
[20] D. B. Araya, J. O. Dabiri, A comparison of wake measurements in motor-
driven and flow-driven turbine experiments, Experiments in Fluids 56
445 (2015) 150.

RI
[21] L. Vermeer, J. Srensen, A. Crespo, Wind turbine wake aerodynamics,
Progress in Aerospace Sciences 39 (6–7) (2003) 467–510.
[22] L. P. Chamorro, F. Porté-Agel, A Wind-Tunnel Investigation of Wind-

SC
Turbine Wakes: Boundary-Layer Turbulence Effects, Boundary-Layer Me-
450 teorology 132 (1) (2009) 129–149.
[23] K. Mahesh, The Interaction of Jets with Crossflow, Annual Review of Fluid

U
Mechanics 45 (2013) 379–407 First.
[24] I. Paraschivoiu, Wind Turbine Design With Emphasis on Darrieus Con-

455 2002.
AN
cept, Presses internationales Polytechnique, Quebec, ISBN 9782553009310,

[25] R. E. Sheldahl, P. C. Klimas, Aerodynamic characteristics of seven symmet-

M
rical airfoil sections through 180-degree angle of attack for use in aerody-
namic analysis of vertical axis wind turbines, Tech. Rep., Sandia National
Laboratories, Albequerque, 1981.
D

460 [26] S. Shamsoddin, F. Porté-Agel, Large eddy simulation of vertical axis wind
turbine wakes, Energies 7 (2) (2014) 890–912, ISSN 19961073.
TE

[27] V. Kumar, M. Paraschivoiu, I. Paraschivoiu, Low Reynolds number vertical

axis wind turbine for Mars, Wind Engineering 34 (2010) 461–476.

[28] R. B. Noll, N. D. Ham, Dynamic Stall of Small Wind Systems, Tech. Rep.,
Aerospace Systems Inc, Burlington, MA, USA, 1983.
EP

465

[29] N. D. Ham, Aerodynamic loading on a two-dimensional airfoil during dy-

namic stall, AIAA Journal (1968) 1927–1934.

[30] S. Shamsoddin, F. Porté-Agel, A large-eddy simulation study of vertical

C

axis wind turbine wakes in the atmospheric boundary layer, Energies 9 (5)
470 (2016) 1–23, ISSN 19961073.
AC

25
 ACCEPTED MANUSCRIPT

Highlights
• The wake of a vertical axis wind turbine is studied in a boundary layer wind tunnel.
• Stereo particle image velocimetry is performed on multiple vertical crosswind planes.
• Turbulence statistics and mean properties of the flow are laterally asymmetric.
• Two pairs of counter rotating vortices are observed and play an important role.

PT
• An analytical model is developed to understand the mechanism behind the vortices.

RI
U SC
AN
M
D
TE
C EP
AC