energies

Article
Net Power Coefficient of Vertical and Horizontal
Wind Turbines with Crossflow Runners
Toni Pujol 1, * ID , Albert Massaguer 1 ID
, Eduard Massaguer 2 , Lino Montoro 1 and
Martí Comamala 1
1 Department of Mechanical Engineering and Industrial Construction, University of Girona,
c/Universitat de Girona 4, 17003 Girona, Spain; albert.massaguer@udg.edu (A.M.);
lino.montoro@udg.edu (L.M.); marti.comamala@udg.edu (M.C.)
2 Nabla Thermoelectrics, c/Llibertat 71, 17820 Banyoles, Spain; eduard@nablatherm.com
* Correspondence: toni.pujol@udg.edu; Tel.: +34-972-418-865

Received: 21 November 2017; Accepted: 27 December 2017; Published: 3 January 2018

Abstract: The feasibility of using crossflow runners as single rotors in vertical-axis wind
turbines (VAWT) or as blades in horizontal-axis wind turbines (HAWT) is numerically studied.
A computational fluid dynamics model is validated from data obtained in a wind tunnel. Three
crossflow runners with different number of blades are tested. Values of drag, lift and torque
coefficients are numerically obtained at different turning velocities. Power coefficients Cp for crossflow
VAWT and HAWT are calculated for different tip-speed ratios (TSR) and runner spin ratios (α). Since
crossflow HAWT consume electrical energy for spinning the runners, the net power coefficient is
estimated. Simulations indicate that a crossflow runner as a single rotor in VAWT should have a high
solidity and work at low TSR. Crossflow runners working as blades in HAWT may achieve low drag
to lift ratios but the Cp is penalized by the amount of energy required for spinning the runners. The
optimum working condition of crossflow HAWT is located within a narrow band of low TSR and α
reaching Cp values < 0.2 only.

Keywords: crossflow runner; crossflow turbine; Magnus turbine; HAWT; VAWT

1. Introduction
Wind energy technologies have experienced an important evolution over the last decades [1].
In large scale designs, the horizontal-axis wind turbine (HAWT) with a three-blade rotor offers
a high aerodynamic performance, being the most suitable option for wind farms with large installed
capacities [2,3]. This type of HAWT commonly starts working at wind speeds over U = 3.5 m s−1 and
with tip-speed ratios (TSR) in the range of 10–13 [2], where TSR is the ratio of the blade tip tangential
velocity Ut to the wind speed U,
ΩR
TSR = , (1)
U
since Ut = Ω·R with Ω the rotor angular velocity and R the blade radius.
Different designs of small- and medium-sized wind turbines can be found with the purpose
of working at other ranges of TSR values [4,5]. It is expected that designs with low cut-in speeds
correspond to those with low TSR. Starting conditions with weak winds have the advantage of
providing more working hours per day, since the most repeatable wind resource is found at low
speeds [6]. These types of wind turbines specifically adapted to small power generation (<10 kW)
become appropriate for urban areas and motorways (see, e.g., [7,8]).
Recent studies have focused on developing small-scale systems that work in the range of very low
TSR (<3). Table 1 reports values of the maximum power coefficient (Cp , the ratio of the power extracted
by the turbine to the power of the incoming flow) achieved in different turbine types working at low

Energies 2018, 11, 110; doi:10.3390/en11010110 www.mdpi.com/journal/energies

Energies 2018, 11, 110 2 of 24

TSR [9–16]. Except [9,12,13], data shown in Table 1 have been obtained experimentally, with some
designs offering cut-in wind speeds as low as 1.6 m s−1 .

Table 1. Example of maximum power coefficients Cp of different turbine types.

TSR Cp Type Turbine Analysis Ref.

0.3 0.45 VAWT Crossflow runner Numerical [9]
0.5 0.30 VAWT Crossflow runner Field testing of prototype [10]
0.7 0.47 VAWT Savonius–Darrieus Wind tunnel [11]
0.8 0.26 VAWT Savonius Numerical [12]
2.6 0.31 VAWT Darrieus Numerical [13]
0.6 0.17 HAWT Multi blade Wind tunnel [14]
1.0 0.35 HAWT Magnus Analytical based on wind tunnel data [15]
2.7 0.40 HAWT 4 blades NACA2404 Wind tunnel [16]

Starting conditions for low wind velocities can be obtained with Magnus type turbines [17]
that consist in HAWT with rotating cylinders instead of standard blades. The Magnus effect arises
when a spinning cylinder is immersed in a flow and it provides a lift force that is the responsible
for producing the required torque (see, e.g., [18]). Several versions of Magnus type turbines have
been investigated, varying the number of spinning cylinders and their rotational speeds ω [15,18–20].
The performance of Magnus type turbines depends on the cylinder spin ratio α, defined as

ωD/2
α= , (2)
U
with D the diameter of the spinning cylinders.
Sedaghat [15] has recently developed a theory for designing Magnus wind turbines with rotating
cylinders working as blades in HAWT, estimating an optimum of Cp = 0.35 at TSR = 1 in the range
of 1.5 < α < 2.5 (see Table 1). In comparison with plain rotating cylinders, higher values of the lift
coefficient have been obtained by using spiral fins [21,22], although the lift coefficient substantially
decreases as the TSR tends to zero [22]. A drawback of rotating cylinders (both plain and with
spirals) is the high value of the drag to lift ratio, which contrasts with the very low ratio recently
obtained with a circulating airfoil specifically designed for providing a Magnus force in HAWT [23].
However, the complexity of the previous mechanism motivates the investigation of simpler options
for applications in wind energy as, for example, the use of crossflow runners.
Once a crossflow runner is immersed in a flow, it achieves an autorotation regime (or free spin
regime αf ). The air flow is able to spin the runner from the fully stopped condition (α = 0) to the free
regime (α = αf ) by varying an external load (that extracts mechanical energy) applied to the shaft of the
runner. Thus, the crossflow runner may be used as a single rotor in vertical-axis wind turbine (VAWT)
when working in the αf ≤ α ≤ 0 range (see Table 2 and Figure 1). On the other hand, external energy
(not related with the incoming air flow) is required for spinning the runner at fixed angular velocities
in other regimes (e.g., 0 < α). This is a proper functioning for crossflow runners that work as blades in
HAWT (as in Magnus HAWT [15]; see Table 2 and Figure 1). Table 2 also indicates the sign convention
of α and the sense of rotation of the crossflow runner in all the figures of the present paper.

Table 2. Regimes of the crossflow runner.

Regime Rotation * External Energy Consumed Use

α < αf anticlockwise Yes -
αf ≤ α ≤ 0 anticlockwise No Single rotor in VAWT
0<α clockwise Yes As blades in HAWT
* Sense of rotation in all the figures of the present paper (e.g., anticlockwise implies ω < 0 in Equation (2)).
Energies 2018,11,
2018,
Energies 11,110
110 3 of3 24
of 24
Energies 2018, 11, 110 3 of 24

Figure 1. Crossflow
Figure runner
runnerapplications: (a)asas asingle
single rotor vertical-axis
in vertical-axis wind turbines (VAWT)
Figure 1.1.Crossflow
Crossflow runner applications: (a)
applications: (a) as aa single rotor
rotor in
in vertical-axiswind
windturbines
turbines(VAWT)
(VAWT)(ω(ω
(ω achieved bythe
theeffect
effectofofthe
theairairflow);
flow); and (b) as blades in horizontal-axis wind turbines (HAWT)
achieved by the effect of the air flow); and (b) as blades in horizontal-axis wind turbines (HAWT)(ω(ω
achieved by and (b) as blades in horizontal-axis wind turbines (HAWT)
(ω must
must be externally
be externallysupplied
supplied by,by,for for example,
example, electrical
electrical motors).
motors).
must be externally supplied by, for example, electrical motors).

Here, the feasibilityofofusing

Here,
Here,the
thefeasibility
feasibility of usingcrossflow
crossflow runners
crossflow inin
runnersin
runners VAWT
VAWT
VAWT (as(as
(as aa single vertical
a single
single rotor)
vertical
vertical as well
rotor)
rotor) as as
as well asininas
well
HAWT
inHAWT
HAWT(as (as blades)
(asblades) is analyzed.
blades)isisanalyzed.
analyzed.TheThe flow-process
Theflow-process
flow-process diagram
diagram of the current work follows Figure 2. 2.
diagram of of
thethe current
current work
work follows
follows Figure
Figure 2.
Section 2 describes the experimental set up. The two-dimensional (2D) Computational Fluid
Section 2 describes
Section the experimental
2 describes set up.
the experimental setThe
up.two-dimensional (2D) Computational
The two-dimensional Fluid Dynamics
(2D) Computational Fluid
Dynamics (CFD) model is explained in Section 3 and its validation is carried out in Section 4.
Dynamics
(CFD) model(CFD) model isinexplained
is explained Section 3in Section
and 3 and its validation
its validation is carried
is carried out out in
in Section 4. Section 4.

Figure 2. Simplified flow-process diagram of the present work.

Figure2.2.Simplified
Figure Simplified flow-process diagram
diagramof
ofthe
thepresent
presentwork.
work.
Energies 2018, 11, 110 4 of 24
Energies 2018, 11, 110 4 of 24

validated,the
Once validated, the2D
2DCFD
CFD model
model is applied
is applied for analyzing
for analyzing the performance
the performance of a crossflow
of a crossflow runner
runner
used inused
a VAWT,in a VAWT,
which meanswhich working
means working
in the αinf ≤the
α ≤αf0≤ α ≤ 0
regime regime
(Section(Section
5). 5).
Values Values
of of
torquetorque
(Cm )
(C ) and power (C ) coefficients are obtained and compared with those corresponding
and power (Cp ) coefficients are obtained and compared with those corresponding to other types of
m p to other types
of turbines.
turbines. ThisThis
casecase has already
has already been analyzed
been analyzed numerically
numerically by Dragomirescu
by Dragomirescu [9] for crossflow
[9] for a bigger a bigger
crossflow
runner. In runner.
comparisonIn comparison with Dragomirescu
with Dragomirescu [9], we provide [9], we provide
detailed detailed information
information of the behavior of the
of
behavior of a single blade and study the consequences of: (1) using flows
a single blade and study the consequences of: (1) using flows with different Reynolds numbers; and with different Reynolds
numbers;
(2) changing andthe(2)orientation
changing the orientation
of the blades. of the blades.
Besides, we also extend the study carried out in Dragomirescu [9] by analyzing the feasibility of
using crossflow runners for working as blades in HAWT. The 2D CFD model is also applied in this
configuration, which
configuration, which means
means working
workingin inthe
the00 << αα regime
regime (Section
(Section 6).
6). In
In this
this case,
case, lift
lift (C
(Cll) and drag (Cdd)
coefficients
coefficients of the spinning cylinders are calculated and used in a 3D analytical approximation (based
on [15]) for estimating the Cpp value as a function of TSR and α. α. In addition, we develop a formulation
for estimating the net power coefficient that takes into account the power needed for spinning the
runners in HAWT.
HAWT. Finally,
Finally,Section
Section77contains
containsthethemain
mainconclusions
conclusionsofofthe thepresent
presentpaper.
paper.

2. Experimental Set Up
Experimental data are obtained with the closed circuit wind tunnel of the Energy Laboratory at
the University
UniversityofofGirona. Figure
Girona. 3 shows
Figure a picture
3 shows of the facility.
a picture of the The windThe
facility. tunnel has tunnel
wind total dimensions
has total
of 7300 mm ×
dimensions of 3600
7300mm mm× × 1800
3600 mm
mm (length
× 1800 mm × height × width)
(length × heightwith a test section
× width) made
with a test of transparent
section made of
glass of dimensions
transparent glass of395 mm × 395
dimensions mm
395 mm(cross-section) and 1325 mmand
× 395 mm (cross-section) (length).
1325 The
mm ducts of the
(length). Thevertical
ducts
wind
of thetunnel arewind
vertical madetunnel
of galvanized
are madesteel
ofwith internal steel
galvanized cornerwith
vanes. The settling
internal cornerchamber
vanes. Theprior to the
settling
contraction
chamber priorcone tocontains an aluminum
the contraction honeycomb
cone contains and a screen
an aluminum to straighten
honeycomb andthe flow direction.
a screen The
to straighten
contraction cone isThe
the flow direction. wooden made and
contraction coneitisiswooden
internally fixed
made andtoitthe outer panels.
is internally fixedThe design
to the outerofpanels.
these
elements
The designfollows theelements
of these technical follows
note of Mehta and Bradshaw
the technical [24]. An
note of Mehta axial
and flow fan [24].
Bradshaw (HMAn 80 T 4 model,
axial flow
Casals,
fan (HM Camprodon,
80 T 4 model,Spain) is controlled
Casals, with a variable-frequency
Camprodon, Spain) is controlled drive
with(Altivar 312 model, Schneider
a variable-frequency drive
Electric,
(Altivar Rueil-Malmaison,
312 model, Schneider France;Electric,
0–50 Hz),Rueil-Malmaison,
which provides a France;
maximum speed
0–50 Hz),of 40 m s−provides
which 1 at the exit
a
of the contraction
maximum speed of cone.
40 m s−1 at the exit of the contraction cone.

Figure 3.
Figure 3. Vertical closed circuit
Vertical closed circuit wind
wind tunnel
tunnel used
used for
for the
the measurements.
measurements. TheThe air
air circulates
circulates clockwise.
clockwise.
The fan is the blue element at the top. The channel test is located inside the area surrounded
The fan is the blue element at the top. The channel test is located inside the area surrounded by
by curtains
curtains
with the with theImage
Particle Particle Image Velocimetry
Velocimetry (PIV)atsystem
(PIV) system at the front.
the front.

The blades of the crossflow runner are manufactured by bending a 0.2 mm thick sheet of steel.
The blades of the crossflow runner are manufactured by bending a 0.2 mm thick sheet of steel.
Its inner edge is situated at 24 mm from the center of rotation, with an outer diameter D = 62 mm and
Its inner edge is situated at 24 mm from the center of rotation, with an outer diameter D = 62 mm and
a transversal length to the free stream air equal to L = 180 mm (see Figure 4). The crossflow turbine is
a transversal length to the free stream air equal to L = 180 mm (see Figure 4). The crossflow turbine
located within the channel test with its rotation axis at a distance of 435 mm from the end of the
contraction cone. Three different configurations of the runner are tested. They differ in the number
of blades, being 6, 11 or 22 (see Figure 4). We point out that the six blades of Prototype 3 are not
Energies 2018, 11, 110 5 of 24

is located within the channel test with its rotation axis at a distance of 435 mm from the end of the
contraction cone. Three different configurations of the runner are tested. They differ in the number of
Energies 2018, 11, 110 5 of 24
blades, being 6, 11 or 22 (see Figure 4). We point out that the six blades of Prototype 3 are not evenly
distributed
evenly since we
distributed built
since weitbuilt
by removing 16 blades
it by removing out ofout
16 blades theof
runner with 22
the runner blades.
with At a given
22 blades. time,
At a given
upstream blades are identified as those that first impact with the incident flow
time, upstream blades are identified as those that first impact with the incident flow (i.e., those(i.e., those situated
on the right-hand
situated side in Figure
on the right-hand side in 4Figure
since 4the direction
since of the incoming
the direction flow Uflow
of the incoming in allUthe figures
in all of the
the figures
present paper is from right to left).
of the present paper is from right to left).

Figure
Figure 4.4. (a)
(a)Three-dimensional
Three-dimensionalview view ofof the
the actual
actual crossflow
crossflow runner;
runner; (b)
(b) its
its two-dimensional
two-dimensional
representation
representationforforthe
theCFD
CFDanalysis (configurations
analysis (configurations of 22, 11 and
of 22, 6 blades,
11 and fromfrom
6 blades, (left)(left)
to (right)); and
to (right));
(c)
anddimensions of theofblade
(c) dimensions (in mm).
the blade (in mm).

Note
Note that
that our
our2D 2Drepresentation
representation does does not not take
take into
into account
accountthe thereinforcement
reinforcement rings rings of
of the
the actual
actual
runner. The rotation axle of the turbine is mounted on a strain gauge wind
runner. The rotation axle of the turbine is mounted on a strain gauge wind tunnel balance (EI400 Series tunnel balance (EI400
Series
model,model,
DELTALAB,DELTALAB, Carcassonne,
Carcassonne, France,France,
resolution resolution
of 0.01 N,of 0.01 N, accuracy
accuracy of ±1%)ofto±1%) measureto measure
lift and
lift and drag
drag forces. forces.
The
The fluid
fluid flow
flow behavior
behaviorisisacquired
acquiredby bymeans
meansof ofaaParticle
ParticleImageImageVelocimetry
Velocimetry(PIV) (PIV)technique.
technique.
We use a 2D PIV system of Dantec Dynamics (Skovlunde,
We use a 2D PIV system of Dantec Dynamics (Skovlunde, Denmark) with a FlowSense Denmark) with a FlowSense camera
camera and
anda
DualPower laser Nd:Yag of 532 nm, 120 mJ/pulse and 14.8 Hz. The
a DualPower laser Nd:Yag of 532 nm, 120 mJ/pulse and 14.8 Hz. The laser is fixed vertically with laser is fixed vertically with a
downwards
a downwards light
lightbeam
beam totobebeable
abletotolight
lightup upa alongitudinal
longitudinalplane planeacross
acrossthethechannel
channeltesttestat
at140
140mm mm
from
from one end of the turbine (in order to avoid the middle reinforcement ring). Seeding particles are
one end of the turbine (in order to avoid the middle reinforcement ring). Seeding particles are
introduced at the end of the channel test by a smoke generator
introduced at the end of the channel test by a smoke generator (Magnum 650 EU model, Martin, (Magnum 650 EU model, Martin,
Arhus,
Arhus, Denmark).
Denmark). We We use
use the
the DynamicStudio
DynamicStudio software software (Dantec
(Dantec Dynamics,
Dynamics, Skovlunde,
Skovlunde, Denmark)
Denmark)for for
configuring,
configuring, acquiring and analyzing data after synchronizing the system with the turning velocity
acquiring and analyzing data after synchronizing the system with the turning velocity
of
of the
the runner.
runner. This
This value
value isis measured
measured with with aa laser
laser tachometer
tachometer(ST-6234B(ST-6234B model,
model, Reed
Reed Instruments,
Instruments,
Wilmington,
Wilmington, NC, USA, resolution of 0.1 rpm, accuracy of ±0.5% + 1 digit). At the sametime,
NC, USA, resolution of 0.1 rpm, accuracy of ±0.5% + 1 digit). At the same time,the theair
air
flow velocity U is independently measured in the channel test with
flow velocity U is independently measured in the channel test with an anemometer (Master 8901 model,an anemometer (Master 8901
model,
Dwyer,Dwyer,
Michigan, Michigan,
IN, USA, IN,resolution
USA, resolution
of 0.01 m of s0.01 m s−1, accuracy
−1 , accuracy of ±2%). of ±2%).
Figure
Figure 5 shows the schematics of the experimental system, wherepoint
5 shows the schematics of the experimental system, where pointAAindicates
indicatesperiodicity.
periodicity.
In
In summary, the experiment consists in: (1) fixing the frequency value in the variable speed drive
summary, the experiment consists in: (1) fixing the frequency value in the variable speed drive
connected
connected to to the
the fan;
fan; (2)
(2) measuring
measuring the the turning
turningvelocity
velocityof ofthe
thecrossflow
crossflow runnerrunner under
under free
free spinning
spinning
conditions;
conditions; (3) (3)synchronizing
synchronizingthe thePIV
PIVsystemsystemaccording
according to to
thethe runner’s
runner’s spinning
spinningvelocity; andand
velocity; (4)
acquiring
(4) acquiringPIVPIVdata, wind
data, tunnel
wind balance
tunnel balance data
dataand andwind
wind speed
speeddata.data.This
Thisprocedure
procedureisisrepeated
repeatedfor for
various wind speed values and configurations of the turbine
various wind speed values and configurations of the turbine (6, 11 and 22 blades). (6, 11 and 22 blades).
Energies 2018,
Energies 2018, 11,
11, 110
110 66 of
of 24
24
Energies 2018, 11, 110 6 of 24

Figure
Figure 5.
5. Control
Control diagram
diagram of
of the
the experimental set up.
experimental set up.
Figure 5. Control diagram of the experimental set up.
3. Simulation Set Up
3.
3. Simulation
Simulation Set Set Up
Up
Several researchers have applied the CFD technique for studying the interaction of flows with
Several
Several researchers have
researchers haveapplied
appliedthe theCFD CFD technique for studying the the interaction of flows
rotating blades [25–31]. Here, simulations aretechnique
carried foroutstudying interaction
with ANSYS-Fluent of flows
(ANSYS, with
Inc.,
with rotating
rotating bladesblades [25–31].
[25–31]. Here, Here, simulations
simulations are are carried
carried out out with
with ANSYS-Fluent
ANSYS-Fluent (ANSYS,
(ANSYS, Inc.,
Inc.,
Canonsburg, PA, USA) in a configuration very similar to that successfully employed in [9]. This
Canonsburg,
Canonsburg,CFD PA, USA)
PA,software in
USA) inhas a configuration
a configuration very similar
very similar to that successfully employed in [9].
commercial been widely applied in thetosimulation
that successfully employed
of air flows in [9].cases
in complex This
This commercial
commercial CFD CFD software
software has has been
been widely widely
applied applied insimulation
the simulation
of air of air flows in complex
involving wind turbines with remarkable success [30].inForthe
simplicity, simulationsflows
are in complex
done cases
in 2D along
cases involving
involving wind wind turbines
turbines with with remarkable
remarkable success success
[30]. [30].
For For simplicity,
simplicity, simulations
simulations are are
done done
in 2D in 2D
along
the plane that is experimentally analyzed with the PIV technique. Air properties are density ρ = 1.225
along
them the that
plane planeis that is experimentally analyzed
experimentally with the PIV technique. Air properties areρdensity
kg −3 and absolute viscosity µ=analyzed
1.7894 × 10 with
−5 Pathe
s. PIV technique. Air properties are density = 1.225
ρkg=m1.225
−3 and m−3 andviscosity
kgabsolute absoluteµviscosity
= 1.7894 µ = 1.7894
× 10 into
−5 × 10 −5 Pa s.
Pa s.two parts: a static domain and a rotating one that
The computational domain is divided
The
The computational
computational domain
domain is
is divided
divided into
into two parts:
two inner aa static
static domain
parts:boundaries domain and aa rotating one
one that
contains the blades of the turbine (see Figure 6). The of theandstatic rotating
domain coincide that
contains
contains the blades of the turbine (see Figure 6). The inner boundaries of the static domain coincide
with the the blades
outer of the turbine
boundaries of the (see Figure
rotating 6). The whereas
domain, inner boundaries of the
the external static domain
boundaries coincide
of the static
with
with the
theouter
outer boundaries
boundaries of the
of rotating
the domain,
rotating whereas
domain, the external
whereas the boundaries
external of the static
boundaries of domain
the static
domain correspond to the limits of the channel test. A transient simulation with a sliding mesh
correspond
domain to the limits
correspond to of the
the channel
limits of test.
the A transient
channel test. simulation
A transient with a slidingwith
simulation mesh a technique
sliding meshfor
technique for the rotating domain is adopted.
the rotatingfor
technique domain is adopted.
the rotating domain is adopted.

Figure 6.
Figure Computational domain
6. Computational domain (in
(in mm).
mm). The
The rotating
rotating domain
domain has
has aa diameter
diameter of
of 70
70 mm.
mm.
Figure 6. Computational domain (in mm). The rotating domain has a diameter of 70 mm.
The mesh
mesh is
is created
createdwithwithANSYS-Meshing
ANSYS-Meshing(ANSYS, (ANSYS, Inc., Canonsburg,
Inc., Canonsburg, PA,PA,USA).
USA). Three
Threelayers of
layers
of The mesh (since
quadrilaterals
quadrilateralsis created
(since with
thethemesh ANSYS-Meshing
meshis is
2D)2D)are
areused (ANSYS,
usedaround
aroundthe Inc., Canonsburg,
theblades
bladeswith PA, USA).
withtriangles
triangles Three
ininthe layers
the rest
rest of of
of the
quadrilaterals (since
computational domain
computational the
domain(see mesh is
(seeFigure 2D)
Figure7). are
7).The used
Theedge around
edgesize the
sizeofofthe blades
theelements with
elementsatatthe triangles
theblades in the
bladesis is0.05 rest
0.05mm of
mmwith the
with a
computational
growth
a growth rate
rate domain
equal
equal to to (see
1.2. The
1.2. Figure
maximum
The 7).
maximum The edge
length size
of the
length of the
edges
of the elements
edgesof the at the
elements
of the elementsblades
is equal is 0.05
to 0.5
is equal mm
to mm with
0.5 mm ina
in the
growth rate
rotating equal
domain andto 51.2.
mm The
in maximum length
the static one. of the
A total edgesofof284,012
amount the elements
elements is equal
is neededto 0.5for
mm in the
meshing
rotating domain and 5 mm in the static one. A total amount of 284,012 elements is needed for meshing
Energies 2018, 11, 110 7 of 24

Energies 2018, 11, 110 7 of 24

the rotating domain and 5 mm in the static one. A total amount of 284,012 elements is needed for
the 2D domain.
meshing the 2DFor this mesh,
domain. the mesh,
For this y+ values
the on
y+ the blades
values are calculated
on the blades are with the function
calculated with thespecifically
function
defined for this purpose in the CFD-Post software. The maximum value of y+ (y+max) for the U = 10 m
specifically defined for this purpose in the CFD-Post software. The maximum value of y+ (y+max ) for
s−1 case
the U = varies
10 m s−from
1 case 1.79 for from
varies a turbine
1.79 rotating at 1500
for a turbine rpm at
rotating anticlockwise to 0.99 rotating
1500 rpm anticlockwise at 750
to 0.99 rpm
rotating
clockwise. Weaker air free stream conditions lead to smaller values of y+max. For example, for U = 2 m
at 750 rpm clockwise. Weaker air free stream conditions lead to smaller values of y+max . For example,
s−1, U
for y+=
max ranges
2 m s−1 , from
y+max0.58 for from
ranges a turbine rotating
0.58 for at 500
a turbine rpm anticlockwise
rotating to 0.33 rotating
at 500 rpm anticlockwise at 200
to 0.33 rpm
rotating
clockwise. Two coarser meshes are also simulated for studying the sensitivity of the results
at 200 rpm clockwise. Two coarser meshes are also simulated for studying the sensitivity of the results to the mesh
size,
to theasmesh
detailed
size,inasthe next section.
detailed in the next section.

7. Detail
Figure 7. Detailofofthe
themesh
meshinin
thethe
interface between
interface the the
between rotating and and
rotating the static regions
the static (left); (left);
regions and detail
and
of the mesh
detail of the around one blade
mesh around one (right).
blade (right).

The transient
The transient numerical
numerical model modeluses usesaatime timestepstepΔt∆tthat thatprovides
provides a turn
a turnof of
1° of
1◦ the turbine
of the for
turbine
anyany
for value
value of the angular
of the angular velocity
velocity(Δt (∆t
= 1/(6N)
= 1/(6N) withwith
N the N turbine’s
the turbine’sturning velocity
turning in rpm).
velocity For
in rpm).
comparison purposes, simulations applying a Δt equivalent to a
For comparison purposes, simulations applying a ∆t equivalent to a 0.2 turn of the 22 blades turbine 0.2° ◦turn of the 22 blades turbine
were carried
were carried out out for
for different
different values
values of of NN and
and results
results did
did notnot significantly
significantly varyvary (changes
(changes in in forces
forces less
less
than 0.5%). The numerical algorithm solves the momentum and mass conservation equations witha
than 0.5%). The numerical algorithm solves the momentum and mass conservation equations with
adouble
doubleprecision
precisionpressure-based
pressure-basedalgorithmalgorithmand andsecond
secondorderorderdiscretization
discretization schemes.
schemes. The The shear
shear stress
stress
transport (SST)
transport (SST) k-ω
k-ω turbulent
turbulent model model is is chosen,
chosen, as as successfully
successfully done done in in [23]
[23] for
for similar
similar conditions,
conditions, and and
the pressure-momentum coupling uses the pressure implicit with
the pressure-momentum coupling uses the pressure implicit with splitting of operator (PISO) method. splitting of operator (PISO) method.
Differences of
Differences of our
our numerical
numerical set set up
up with
with [9][9] are
are that,
that, here:
here: (1)
(1) wewe adapt
adapt the
the time
time step
step toto have
have thethe same
same
angle of
angle ofrotation
rotationper perouter
outeriteration
iteration forfor
all all
thethe
casescases analyzed;
analyzed; (2) flow
(2) our our flow
regime regime corresponds
corresponds to lowerto
lower Reynolds numbers (Re = UDρ/µ ), as discussed in Section 5;
Reynolds numbers (Re = UDρ/µ), as discussed in Section 5; and (3) we use the SST k-ω model as in [23] and (3) we use the SST k-ω model
as in [23]
instead ofinstead of the realizable
the realizable k-ε model k-ε sincemodel since it isto
it is expected expected
be moretorobustbe moreandrobust and accurate.
accurate.
The boundary condition at the inlet is defined as
The boundary condition at the inlet is defined as a uniform velocity with a 5%a uniform velocity with a 5% ofof turbulence
turbulence
intensity and a turbulent viscosity ratio equal to 10 for all cases,
intensity and a turbulent viscosity ratio equal to 10 for all cases, both being typical values for both being typical values for these
these
turbulence parameters.
turbulence parameters.AAlower lowervaluevalueofofturbulence
turbulence intensity
intensity (=1%)
(=1%) at the inlet
at the waswas
inlet alsoalso
tested for the
tested for
U = 2 m s −1 and−α 1 = 0.48 case. Results deviated less than 0.7% from
the U = 2 m s and α = 0.48 case. Results deviated less than 0.7% from those obtained with a 5% those obtained with a 5% turbulence
intensity. Aintensity.
turbulence pressure equal to 0 Paequal
A pressure (relative)to 0 isPafixed at the is
(relative) outlet
fixedboundary.
at the outletAll other boundaries
boundary. are
All other
stated as smooth
boundaries walls.
are stated as smooth walls.
We apply the criterion
We apply the criterion of of performing
performing aa maximum maximum of of 30
30 inner
inner iterations
iterations atat each
each time
time step.
step. InIn some
some
cases, the
cases, the convergence
convergence criterion
criterion (residuals
(residuals less less than
than10 10 )) is
−−4
4 is achieved
achieved earlier.
earlier. AA minimum
minimum amountamount of of
1044 time steps is simulated. Values of forces and torque are monitored for the whole crossflow runner
10 time steps is simulated. Values of forces and torque are monitored for the whole crossflow runner
(all blades)
(all blades) as as well
well as as for
for one
one individual
individual blade. blade. After
After an
an initial
initial time
time span,
span, the
the behavior
behavior of of forces
forces and
and
torque shows
torque shows aa periodic
periodic signal.
signal. Once
Once we we confirm
confirm thatthat the
the system
system reaches
reaches aa stationary
stationary behavior
behavior in in the
the
sense that the oscillatory signal for one individual blade does not
sense that the oscillatory signal for one individual blade does not remarkably change when shifted remarkably change when shifted
360°
360 ◦ in
in time,
time, values
valuesofofboth bothx xand andy yforces
forces asaswell
wellas as
torque
torque arearerecorded.
recorded. Since the the
Since simulation
simulation is 2D,is
the values
2D, the values of forces and and
of forces torque reported
torque reported by theby solver are per
the solver are unit length,
per unit being
length, multiplied
being multipliedby Lby(the L
length of the turbine) for carrying out the comparison
(the length of the turbine) for carrying out the comparison with experimental data.with experimental data.
Forces Fx (positive leftwards) and Fy (positive downwards) (see the inset of the CFD simulations
box in Figure 2) are also used for calculating the aerodynamic coefficients of lift Cl and drag Cd, as
usual [23]
 Energies 2018, 11, 110 8 of 24

𝐹𝑦
𝐶𝑙 = , (3)
𝑞𝑆𝑟𝑒𝑓
Energies 2018, 11, 110 𝐹𝑥 8 of 24
𝐶𝑑 = , (4)
𝑞𝑆𝑟𝑒𝑓
where
Forcesq F=x½(positive
ρU2 is theleftwards)
kinetic energy
and Fper unit volume of the incoming flow and Sref is the reference
y (positive downwards) (see the inset of the CFD simulations
area of the crossflow turbine (Sref = D·L).
box in Figure 2) are also used for calculating the aerodynamic coefficients of lift Cl and drag Cd ,
The torque coefficient Cm is calculated from [9]
as usual [23]
F𝑀y
𝐶𝑚
Cl== , , (5) (3)
𝑞𝑆qS
𝑟𝑒𝑓 𝐷/2
re f
where M is the torque acting on the blades (value reported
Fx by the solver multiplied by L) with respect
to the center of rotation of the runner. Cd = , (4)
qS re f

q = 12Validation
4. CFD
where ρU2 is the kinetic energy per unit volume of the incoming flow and Sref is the reference
area of the crossflow
Here, turbine
the purpose is to(Sassess ·L).ability of the CFD model to correctly reproduce the observed
ref = Dthe
The torque
trend of bothcoefficient
lift and dragCmforces.
is calculated
Since thefrom [9]
experimental set up does not include any external brake
or power for modifying the rotating speed of the runner, laboratory data correspond to the free
M
regime αf (or autorotation) case. Autorotation Cm =cannot be observed
, in a plain cylinder but it arises in (5)
other types of rotors as, for example, the Savonius qSre f D/2
one [26] and the cycloidal propeller [27,28].
The angular velocity Nf (in terms of rpm) of the free regime achieved in our crossflow runners
where M is the torque acting on the blades (value reported by the solver multiplied by L) with respect
as a function of the incoming wind speed U is shown in Figure 8. Note that values of Nf are defined
to the
as center of rotation
negatives of theturn
since runners runner.
anticlockwise.
The free turning velocity obtained with high to medium solidity levels (11 and 22 blades) is
4. CFD Validation
almost the same in the range of U < 15 m s−1 as it is seen in Figure 8 (changes less than 6%). For higher
wind
Here,speeds, the free
the purpose turning
is to assessvelocity depends
the ability of theonCFD
the model
soliditytoofcorrectly
the runner, decreasing
reproduce the as the
observed
trendnumber
of bothoflift
blades
and decreases.
drag forces.TheSince
crossflow runner with the
the experimental setlargest
up doessolidity keeps a any
not include linear relationship
external brake or
powerbetween Nf and U the
for modifying (linear regression
rotating speedwith a correlation
of the coefficientdata
runner, laboratory R2 = 0.998). The lowest
correspond to thewind speed α
free regime f
that produces case.
(or autorotation) autorotation is almost
Autorotation the same
cannot for both 22
be observed in and 11 blade
a plain turbines
cylinder (slightly
but it arises inbelow
other4m
types
s−1). As the solidity of the turbine is reduced, the lowest wind speed for achieving the free regime
of rotors as, for example, the Savonius one [26] and the cycloidal propeller [27,28].
increases and, at the same time, the runner turning velocity becomes substantially smaller than in the
The angular velocity Nf (in terms of rpm) of the free regime achieved in our crossflow runners as
other configurations. In an ideal frictionless coupling between the runner shaft and the turbine (i.e.,
a function of the incoming wind speed U is shown in Figure 8. Note that values of Nf aretodefined
no bearing friction forces), the lowest flow velocity that would turn the runner would tend zero.
as
negatives since runners turn anticlockwise.

Figure 8. Free
Figure regime
8. Free Nf N
regime asf as
a function
a functionofofthe
thefree
free stream
stream air velocityUUachieved
airvelocity achievedin in crossflow
crossflow runners
runners
withwith
6, 116,and 22 blades (wind tunnel experimental
11 and 22 blades (wind tunnel experimental data). data).

TheDrag Fx (positive leftwards in Figure 6) and lift Fy (positive downwards in Figure 6) forces for
free turning velocity obtained with high to medium solidity levels (11 and 22 blades) is almost
the free regime conditions of Figure 8 are measured and simulated (see Figures 9 and 10). For U < 15
the same in the range of U < 15 m s−1 as it is seen in Figure 8 (changes less than 6%). For higher
wind speeds, the free turning velocity depends on the solidity of the runner, decreasing as the
number of blades decreases. The crossflow runner with the largest solidity keeps a linear relationship
between Nf and U (linear regression with a correlation coefficient R2 = 0.998). The lowest wind
speed that produces autorotation is almost the same for both 22 and 11 blade turbines (slightly below
4 m s−1 ). As the solidity of the turbine is reduced, the lowest wind speed for achieving the free regime
increases and, at the same time, the runner turning velocity becomes substantially smaller than in
Energies 2018, 11, 110 9 of 24

the other configurations. In an ideal frictionless coupling between the runner shaft and the turbine
(i.e., no bearing friction forces), the lowest flow velocity that would turn the runner would tend to zero.
Energies 2018, 11, 110 9 of 24
Energies Fx (positive
Drag2018, 11, 110 leftwards in Figure 6) and lift Fy (positive downwards in Figure 6) forces for
9 of 24 the

freemregime
s−1 (Nconditions of Figure
f > −1500 rpm), the 8global
are measured and simulated
trend experimentally (see Figures
obtained 9 and 10).
is correctly For U <by15the
captured m s−1
m s−1 (Nf > −1500 rpm), the global trend experimentally obtained is correctly captured by the
> −1500 rpm),
(Nf simulations. the global
However, trendlarge
at very experimentally
values of Uobtained
(on the is correctly
order of 26captured
m s−1, Nfby < the simulations.
−1900 rpm),
simulations. However, at very large values of U (on − the
1 order of 26 m s−1, Nf < −1900 rpm),
However, at verybetween
discrepancies large values of U (on
simulations andthe order of 26become
experiments Nf < −1900
m s , greater. Somerpm), discrepancies
reasons between
that may explain
discrepancies between simulations and experiments become greater. Some reasons that may explain
the over prediction
simulations of the simulation
and experiments values in
become greater. comparison
Some reasonswith
that the
may experimental
explain thedata
overare: (1) the of
prediction
the over prediction of the simulation values in comparison with the experimental data are: (1) the
the finite
finite
effect ofvalues
simulation the actual
effect of the actual
runner geometry
in comparison (endexperimental
with the
runner geometry
plates effect not occurring
data
(end plates effect not are: (1) in
occurring thethe
in
simulations);
finite
the
(2) the
effect of the
simulations); actual
(2) the
effect
runner of the
geometry reinforcement
(end plates rings
effect (this
not geometry
occurring not
in included
the in the
simulations); virtual
(2) the domain
effect ofsimulated);
the and
reinforcement
effect of the reinforcement rings (this geometry not included in the virtual domain simulated); and
rings(3) (this
unequal blades in
geometry theincluded
not actual runner
in the(out of balance).
virtual domain simulated); and (3) unequal blades in the
(3) unequal blades in the actual runner (out of balance).
actual runner (out of balance).

Figure 9. Drag
Figure 9. Dragforces
forces measured
measuredandandsimulated for crossflow
simulated for crossflowrunners
runnerswith:
with:
2222 blades
blades (a);(a); 11 blades
11 blades (b); (b);
Figure 9. Drag forces measured and simulated for crossflow runners with: 22 blades (a); 11 blades (b);
andand
six six blades
blades (c),(c), under
under free
free regime
regime conditions(see
conditions (seeFigure
Figure8).
8).Note
Notethat
thaterror
errorbars
bars are
are smaller
smaller than the
and six blades (c), under free regime conditions (see Figure 8). Note that error bars are smaller than
sizethe
of size of the symbols.
the symbols.
the size of the symbols.

Figure
10.10. Lift
Lift forces
forces measuredand
measured and simulated for crossflow
crossflowrunners
runnerswith: 2222
blades (a);(a);
11 blades (b); (b);
Figure
Figure 10. Lift forces measured andsimulated
simulated for
for crossflow runners with:
with: blades
22 blades 11 blades
(a); 11 blades (b);
and
andand six
six six blades
blades (c),
(c),(c), under
under free
free regime conditions (see Figure 8). Note that error bars may be smaller
blades under freeregime
regimeconditions
conditions (see Figure8).
(see Figure 8).Note
Notethat
thaterror
error bars
bars maymaybe be smaller
smaller
than the size of the symbols.
than thethe
than sizesize
of the symbols.
of the symbols.
Energies 2018, 11, 110 10 of 24

Energies
Results2018,in11,Figures
110 10 ofdata.
9 and 10 include the uncertainty values of both measured and simulated 24

The error bars of data are less than 6% for all cases and may be even lower than the size of the symbol
used inResults
the graphs. The error
in Figures 9 andbars added the
10 include to the simulation
uncertainty values
values correspond
of both measured to and
the grid convergence
simulated data.
index
TheGCIerrorfine21
barsthat is anare
of data estimation
less than of 6%thefor uncertainty
all cases andproduced
may be even bylower
the discretization
than the size of procedure
the symbol [32].
Here, the evaluation of the GCI index for the 22 blades runner is carried
used in the graphs. The error bars added to the simulation values correspond to the grid convergence
fine21 out by using three meshes
that differ
index GCIinfine21
the number
that is anofestimation
elementsof employed: a coarse
the uncertainty mesh with
produced by the 45,362 elements,
discretization a medium
procedure size
[32].
mesh
Here,with 82,585 elements
the evaluation of the GCIand a
fine21 fine
index mesh
for with
the 22 284,012
blades elements.
runner is carried Grid
out patterns
by using of
threecoarse
meshes and
that differ
medium in theare
meshes number of elements
identical than those employed:
for the a coarse mesh with
finer mesh. The45,362 elements,
goal values fora calculating
medium sizethe
mesh with 82,585 elements and a fine mesh with 284,012 elements.
GCIfine21 index are Cd and Ci. CFD results for the 22 blades rotor configuration under seven Grid patterns of coarse and medium
different
meshes with
scenarios are identical
flow inletthan those for ranging
velocities the finer from
mesh.2The m sgoal
−1 to values
26.5 mfor s calculating
−1 and turning thevelocities
GCIfine21 index
ranging
are C and C . CFD results for the 22 blades rotor configuration
from −2358 rpm to 1000 rpm lead to GCIfine21 values below 17% (Ci as a key variable) and below
d i under seven different scenarios with8%
flow inlet velocities ranging from 2 m s −1 to 26.5 m s−1 and turning velocities ranging from −2358 rpm
(Cd as a key variable). Large values of the GCIfine21 index may indicate poor grid designs. However, in
to of
four 1000
therpmseven lead to GCI
cases, thefine21 values
GCIfine21 indexbelow 17% (C
is below i asfor
5% a key variable)
Ci and in six of andthebelow
seven8% (Cd as
cases, a below
it is key
variable). Large values of the GCIfine21 index may indicate poor grid designs. However, in four of the
4% for Cd.
seven cases, the GCIfine21 index is below 5% for Ci and in six of the seven cases, it is below 4% for Cd .
Besides quantitative values, the flow qualitative behavior was experimentally investigated.
Besides quantitative values, the flow qualitative behavior was experimentally investigated.
Figure 11 plots the contours of the air velocity magnitude obtained with the PIV experimental
Figure 11 plots the contours of the air velocity magnitude obtained with the PIV experimental technique
technique for different values of the incoming flow U. Blank regions correspond to those zones with
for different values of the incoming flow U. Blank regions correspond to those zones with no available
no information
available information since they are located in the shaded zone of the laser beam. As pointed out
since they are located in the shaded zone of the laser beam. As pointed out above, the free
above, the free turning velocity
turning velocity Nf for the three Nturbines
f for the three turbines (6, 11 and 22 blades) differs, increasing with
(6, 11 and 22 blades) differs, increasing with U. For a weak
U. air
Forfree
a weak air free stream, N f is low and the flow tends to be symmetrical with low values of lift
stream, Nf is low and the flow tends to be symmetrical with low values of lift and drag forces,
andespecially
drag forces, in low especially in low solidity
solidity runners. runners. due
As Nf increases As Ntof increases
higher values due of to U,
higher values
a region of U,
of low a region
velocity
of low velocity
situated near situated nearside
to the lower to theof thelowerexitside ofrunner
of the the exitisof the runner
formed. This is formed.
region This region
is clearly is clearly
differentiated
differentiated
from the restfrom of thethe restinofthe
wake therunnerwake with in thehigh
runner with
solidity high solidity
(Prototype 1). (Prototype 1).

Figure 11.11.
Figure Velocity
Velocitycontours as aasfunction
contours of Uof(columns)
a function andand
U (columns) turbine
turbinetypes (rows)
types (prototype
(rows) 1, 2 1,
(prototype and
3 =2 and 3 = and
22, 11 22, 116 and 6 blades,
blades, respectively;
respectively; the the blade
blade distributionisis for
distribution for illustrative
illustrativepurposes
purposesonly).
only).
Experimental
Experimental data.
data.

The above behavior agrees with the streamlines obtained from the experimental data shown in
Figure 12. At a fixed value of wind speed U, the flow pattern deviates from symmetry as the number
 Energies 2018, 11, 110 11 of 24

The above behavior agrees with the streamlines obtained from the experimental data shown in
Energies 2018, 11, 110 11 of 24
Figure 12. At a fixed value of wind speed U, the flow pattern deviates from symmetry as the number of
blades increases, especially at the wake. There, a low velocity zone at the lower left exit of the turbine
of blades increases, especially at the wake. There, a low velocity zone at the lower left exit of the
is clearly observed in most of the flow speed U cases for the 22 blades rotor.
turbine is clearly observed in most of the flow speed U cases for the 22 blades rotor.

Figure
Figure 12. 12. Streamlines
Streamlines for several
for several values
values U (columns)
of Uof(columns) and and turbine
turbine types types (rows)
(rows) (prototype
(prototype 1,
1, 2 and
2 and 3 = 22, 11 and 6 blades, respectively; the blade distribution is for illustrative purposes only).
3 = 22, 11 and 6 blades, respectively; the blade distribution is for illustrative purposes only).
Experimental data.
Experimental data.

ForFor comparison
comparison purposeswith
purposes withFigures
Figures11 11 and
and 12,
12, Figure
Figure 13 13 shows
showsthe thevelocity
velocitycontours
contours andand
vectors simulated with ANSYS-Fluent for the U = 10 m s − 1 case under N conditions. In the high
vectors simulated with ANSYS-Fluent for the U = 10 m s case under Nf conditions. In the high
−1 f
solidity configuration at high incoming velocities, there is a zone of low velocities attached to the
solidity configuration at high incoming velocities, there is a zone of low velocities attached to the
lower exit of the turbine (confirming the observation with the PIV technique) that, at the same time,
lower exit of the turbine (confirming the observation with the PIV technique) that, at the same time,
includes a vortex located inside the runner, as already pointed out in [9]. The effects of these features
includes a vortex located inside the runner, as already pointed out in [9]. The effects of these features
are discussed in Section 5.
are discussed in Section 5.
In the upper case of Figure 13 (runner with 22 blades) a region with a significant change in the
In the upper case of Figure 13 (runner with 22 blades) a region with a significant change in the
velocity field is observed in the lowermost part of the runner where opposite directions between the
velocity
blade field is observed
velocity (turning in the lowermost
anticlockwise) and part
theof the runner
incoming flow where opposite
velocity directions
are found. between
For this runner,the
blade velocity
a high (turning
velocity value isanticlockwise)
predicted at theand bladethetip
incoming flow upstream
exit of several velocity are found.
blades and For this runner,
an inner intense a
high velocity value is predicted at the blade tip exit of several upstream blades
anticlockwise vortex is formed (confirming the results reported in [9]). This vortex is caused by and an inner intense
the
anticlockwise
geometry of the blades since their tips point at different directions towards the inner of the runner andthe
vortex is formed (confirming the results reported in [9]). This vortex is caused by
geometry
lead to aofblockage
the blades since
of the their tips
incoming flowpoint
at theat different
upstream directions
lower zone oftowards
the runnerthe(thereby
inner ofgenerating
the runner
andanlead
innertoregion
a blockage
of lowofvelocity).
the incomingAt the flow
loweratleft theexit
upstream
region oflower zone of
the runner, the runner
a series (thereby
of clockwise
generating an inner vortices
and anticlockwise region of low velocity).
is observed. At the lower
This behavior is not left exitreproduced
exactly region of the whenrunner, a series
simulating the of
runner with
clockwise and 11 blades due tovortices
anticlockwise its loweris solidity.
observed. In this
Thiscase, the flow
behavior is that
not crosses
exactly the turbine is not
reproduced when
intensively
simulating theforced
runnerto follow
with 11 anblades
upward direction
due since the
to its lower separation
solidity. between
In this case, the
the blades is larger
flow that thanthe
crosses
in theisrunner
turbine with 22 blades.
not intensively forced to follow an upward direction since the separation between the
Since the simulation
blades is larger than in the runner of the with
high 22
solidity
blades. runner (22 blades) under the free regime condition
satisfactorily reproducesofthe
Since the simulation experimental
the high solidity data for U(22
runner < 15 m s−1 ,under
blades) the methodology
the free regimedescribed in
condition
satisfactorily reproduces the experimental data for U < 15 m s , the methodology described in Section
−1

3 is applied to study the use of crossflow runners (with 22 blades) in VAWT (Section 5) and in HAWT
(Section 6).
Energies 2018, 11, 110 12 of 24

Section 3 is applied to study the use of crossflow runners (with 22 blades) in VAWT (Section 5) and in
Energies 2018, 11, 110 12 of 24
HAWT (Section 6).

Figure 13.Contour
Figure13. Contour(a);(a);and
andvector
vector(b)
(b)plots
plotsofofthe flow
the flowvelocity simulated
velocity simulated Nf N
at at with U = 10 m s−1 and
f with U = 10 m s−1 and
runners
runnerswith
with22,
22,1111and
and6 6blades
blades(from
(from(top)
(top)toto(bottom)).
(bottom)).The
Therange
rangeofofvalues
valuesofofthe
thecolor
colorbar
barfor
forthe
the
six-blade case differs from those of 22 and 11 cases for a better representation.
six-blade case differs from those of 22 and 11 cases for a better representation.

5.5.VAWT
VAWT
AAvertical-axis
vertical-axissingle
singlerotor
rotorcrossflow
crossflowrunnerrunnerworks worksininthe regime
the regime αf α≤f ≤α α≤≤0.0.In Inthis case,TSR
thiscase, TSRisis
equal −
equal to −α (since the turning velocity is negative by convention) and the performance curve ofthe
to α (since the turning velocity is negative by convention) and the performance curve of the
wind
windturbine
turbineisisachieved
achievedby bybraking
brakingititfromfromthe therunaway
runawaycondition
condition(α(α==αα f )f)to
tothe
thefully
fullystopped
stoppedone one
(α(α==0).0).Simulations
Simulationsuse useflow
flowspeeds
speeds 4m 4m s−s1−1
, 5, 5mms−s1−1and
and1010mms−s−1 1 , being above the lower speed for
being above the lower speed for
autorotation
autorotationexperimentally
experimentallyfound foundfor forthe
the2222blades
bladesrunner
runnercase
case(Figure
(Figure8).8).
As Asan anexample
example of the resultsresultsfound,
found,we wediscuss
discussthethe case
case U =U4 =m4s−1mand
of of s−1Nand N =rpm
= −100 −100(Rerpm
= 1.7
(Re = 1.7 × 10 4 ). Under these conditions, the torque coefficient C calculated from Equation (5) for
× 10 ). Under these conditions, the torque coefficient Cm calculated
4 m from Equation (5) for the whole
the whole
runner runner
shows an shows
irregularan oscillatory
irregular oscillatory
behavior with behavior
a peakwith a peak
in the in the frequency
frequency spectrum at spectrum
36.7 Hz,
atwhich
36.7 Hz, which
matches matches
the blade tothe blade
blade to blade
passing passing(=−22
frequency frequency (=−22
N/60 with N N/60
= −100with = −100
rpm;NFigure rpm;
14). The
Figure 14). The behavior
behavior of Cm for the whole of C for the whole rotor reproduces that reported
m rotor reproduces that reported in [9] although here with a higherin [9] although here with
aamplitude
higher amplitude of the oscillations
of the oscillations most due
most likely likely to due to the differences
the differences in shape in shape
and size andofsizetheofblades.
the blades.
For a
For a single blade, the time evolution
single blade, the time evolution of Cm (calculatedof C m (calculated from Equation (5) with M the
from Equation (5) with M the torque acting on onetorque acting on
one blade
blade with with respect
respect to the
to the center
center of rotation
of rotation of the of runner)
the runner) clearly
clearly has ahas a signal
signal with with a frequency
a frequency equal
equal
to thetoturning
the turning velocity
velocity (=−N/60(=−=N/60
0.6 Hz= 0.6
with HzNwith N=
= −100 −100
rpm; rpm;14).
Figure Figure 14).
Energies 2018, 11, 110 13 of 24

Energies 2018, 11, 110 13 of 24

Energies 2018, 11, 110 13 of 24

Figure 14. Cm as a function of time for one blade and for the whole rotor (22 blades) in a VAWT with
Figure
Figure Cm C
14. 14. asm aasfunction
a function
ofoftime
timefor
forone
one blade
blade and
andfor
forthe
thewhole
wholerotor (22 (22
rotor blades) in a VAWT
blades) with Uwith
in a VAWT
U = 4 m s−1 and
−1 N = −100 rpm.
U = =4 4mms−1
s and
and −100
NN= =−100 rpm.
rpm.

The variation of Cm as a function of the blade position is shown in Figure 15, where data
TheThevariation of ofCmCmasasa afunction
variation function of of the
the blade positionisisshown
blade position shownin in Figure
Figure 15, 15, where
where datadata
correspond to thetoblade
correspond the cycle
blade ranging
cycle ranging from
from 6.424
6.424 s to
s to7.024
7.024s in
s in Figure
Figure 14. 14. Angle
Angle = 0 ◦θin
= Figure
0° in Figure
15 15 15
correspond to the blade cycle ranging from 6.424 to 7.024 s in Figure 14. Angle θ = 0° in Figure θ
refers to
refers the
refers uppermost
to
to the the position
uppermost
uppermost position
position ofofof
thetheblade
the blade in
blade Figure
Figure444and
in Figure
in and
and increases
increases clockwise.
clockwise.
increases clockwise.

Figure 15. Cm (radius of the polar graph) as a function of the angular position for a single blade. U = 4
Figure 15. Cm15.
Figure (radius of theofpolar
Cm (radius
m s−1 and N =−1−100
graph)
the polar as aas
graph) function of the
a function
rpm (blade turns anticlockwise).
angular
of the angularposition
position for
for aa single blade. U = 4
single blade.
U = 4 m s and N = − 100 rpm (blade turns
m s and N = −100 rpm (blade turns anticlockwise).
−1 anticlockwise).

The pattern observed in Figure 15 can be explained from the velocity vectors and pressure
The pattern
The pattern observedininFigure
observed Figure 1515 cancan
be explained
be from thefrom
explained velocity
thevectors and pressure
velocity vectors contours
and
contours of Figure 16. A single blade produces a maximum torque once situated ◦in the 210° < θ <pressure
280°
of Figure 16. A single blade produces a maximum torque once situated in the 210 < θ < 280◦ range
contours of Figure 16. A(Figure
range approximately single blade
15). It produces
coincides awith maximum
the regiontorque
whereoncethesituated
positionin ofthe
the210°
blade < θwith
< 280°
approximately (Figure 15). It coincides with the region where the position of the blade with respect
range approximately
respect to the incoming (Figure
flow 15). It coincides
is suitable with thenon-negligible
for generating region where liftthe position
forces of the
(high flow blade
speed andwith
to the incoming flow is suitable for generating non-negligible lift forces (high flow speed and low
respect to the incoming
lowpressure
pressure below
below thetheflow
blade is suitable
blade
and and lowfor
low flow generating
flow
speed speed
and highand non-negligible
above it). lift
high pressure
pressure Thisforces
above (high
it). This
contrasts flow
contrasts
with speed
the drag withand
the drag
low pressure mechanism
mechanism below for generating
the bladetorque
for generating and low torque observed
flowinspeed
observed in
the 10 and
◦ the
< θ <high 10°
◦ < θ <
pressure
40 region 40° region
above 15
(see Figures (see
it).and Figures
This Drag is and
15
16).contrasts with
16).
the drag Drag is also
alsomechanism responsible
responsible of of
forthegeneratingthe main contribution
torquefor
main contribution observed for
creating ain creating
the 10°
negative a negative
< θ <(100
torque ◦ torque
40° <region ◦
θ < 140(see (100° < θ
Figures
in Figure < 140°
15)15 and
16). Drag is also responsible of the main contribution for creating a negative torque (100° < θ(see
in Figure
due to 15)
a due to a quasi-perpendicular
quasi-perpendicular impact of the impact
incoming of
flowthe incoming
on the blade flow
surface on the
(see blade
Figure surface
16). < 140°
Figure 16).
in Figure 15) due to a quasi-perpendicular impact of the incoming flow on the blade surface (see
Figure 16).
Energies 2018, 11, 110 14 of 24
Energies 2018, 11, 110 14 of 24

−1
Figure 16. Velocity
Figure 16. Velocityvectors
vectors(a);
(a);and
andpressure
pressurecontours
contours(b)
(b)for andNN==−−100
forUU= =4 4mms s−1and 100 rpm (turning
4
anticlockwise). Re = 1.7 ××10
anticlockwise). Re 104. .

Typical crossflow
Typical crossflow runners
runners inin hydraulic
hydraulic turbines
turbines areare drag-driven
drag-driven due due toto the
the use
use of
of aa water
water inlet
inlet
conduit that points towards a limited number of blades (becoming an impulse-type
conduit that points towards a limited number of blades (becoming an impulse-type machine). Here, machine). Here,
in contrast,
in contrast, the
the entire
entire crossflow
crossflow runner
runner isis immersed
immersed in in the
the flow
flow (air)
(air) and,
and, from
from the
the above,
above, liftlift is
is the
the
main contributor
main contributor for
for producing
producing the the net
net torque.
torque.
The flow simulation predicts the existence
The flow simulation predicts the existence of of aa vortex
vortex inin the
the lower
lower left
left region
region behind
behind thethe runner,
runner,
whose change
whose changein in intensity
intensity causes
causes signalsignal variations
variations betweenbetween
cycles forcycles
a single for a single
blade blade and,
and, consequently,
consequently,
for for the whole
the whole runner. Althoughrunner.
this Although
behavior isthis behavior
observed forisother
observed
valuesforofother
U andvalues
N inof theU region
and N
αf ≤ α ≤ 0, an intense vortex shedding as that found behind static cylinders at similar Re [33] Re
in the region α f ≤ α ≤ 0, an intense vortex shedding as that found behind static cylinders at similar is
[33]detected.
not is not detected.
The performance
The performance of of the
the runner
runner isis expressed
expressed inin terms
terms ofof the
the power
power coefficient
coefficientCCpp,,
𝑃
𝐶𝑝 = P , (6)
C p = 𝑞𝑈𝑆𝑟𝑒𝑓 , (6)
qUSre f
where P (=Ω·M ) is the power extracted by the turbine.
P (=Ω
whereFrom ·M) is the(1)
Equations power extracted
and (5), by the
𝐶𝑝 = 𝑇𝑆𝑅 𝐶𝑚turbine.
where 𝐶𝑚 is obtained after averaging the simulated
From Equations (1) and (5), C = TSR
values during five runner rotations once the single
p C m where Cmdata
blade is obtained after averaging
clearly indicates thesignal.
a periodic simulated
The
values during five runner rotations once the single blade data clearly indicates
results of Cp are shown in Figure 17 where the values are substantially lower than those exposed a periodic signal.
in
The results of Cp[9].
Dragomirescu areThis
shown in Figureour
is because 17 where
turbine thehas
values are substantially
a higher aspect ratiolower than those
(=L/(D/2)) and exposed
a lower
in Dragomirescu
Reynolds number,[9]. This
and is because
both effects areour turbine
known to has a higher
reduce aspect
efficiency inratio
VAWT (=L/(D/2))
[34]. The and a lower
comparison
Reynolds
with othernumber,
types ofand bothturbines,
wind effects are known
such to reduce efficiency
as Savonius, horizontalinmulti-blade
VAWT [34]. and
The horizontal
comparisonthree-
with
other types of wind turbines, such as Savonius, horizontal multi-blade and horizontal
blade, confirms that crossflow runners working as VAWT may only be suitable solutions for very three-blade,
confirms that crossflow
low TSR values (Figure runners
17). working as VAWT may only be suitable solutions for very low TSR
values (Figure 17).
The simulations here shown do not include any resistant torque provided by shaft bearings.
Therefore, positive values of Cm are reported at turning velocities beyond the runaway points
experimentally obtained. Thus, crossflow runner data of Cp in Figure 17 correspond to upper bounds.
We also note that our study has focused on VAWT with straight blades. However, twisted blades
may enhance the power output, as already pointed out in detailed 3D numerical simulations of
Savonius [35] and Darrieus rotors [36]. In addition, twisted blades substantially reduce the oscillations
of torque and power, although the interactions with the vortices at the wake avoid the existence
of completely smooth output signals [36]. On the other hand, curved blades along the z-direction
contribute to alleviate the bending moments of straight blades, although oscillations of torque and
power similar to those of the straight blade case are also expected [36].

Figure 17. Power coefficient Cp as a function of TSR for crossflow runners (symbols) working as rotors
in VAWT and comparison with other types of turbines.
results of Cp are shown in Figure 17 where the values are substantially lower than those exposed in
Dragomirescu [9]. This is because our turbine has a higher aspect ratio (=L/(D/2)) and a lower
Reynolds number, and both effects are known to reduce efficiency in VAWT [34]. The comparison
with other types of wind turbines, such as Savonius, horizontal multi-blade and horizontal three-
blade, confirms
Energies 2018, 11, 110that crossflow runners working as VAWT may only be suitable solutions for
15very
of 24
low TSR values (Figure 17).

Figure 17.
17. Power coefficient C
Cp as a function of TSR for crossflow runners (symbols) working as rotors
Figure Power coefficient p as a function of TSR for crossflow runners (symbols) working as rotors
in VAWT and comparison with other
in VAWT and comparison with other types typesof
ofturbines.
turbines.

The effect of modifying the orientation of the blades has also been investigated. Three rotors, each
one with 22 blades of equal shape to those used in Figure 4 but inclined to form a blade tip exit angle
of 20.5◦ , 30.5◦ and 35.5◦ (the nominal blade tip exit angle is 25.5◦ ; see Figure 4c), have been simulated
at Re = 1.7 × 104 and different TSR. Results are shown in Table 3. For all the cases analyzed, the power
coefficient is maximum when using the 30.5◦ configuration, with a 12% increase in comparison with
the 20.5◦ case (TSR = 0.08, turning velocity equal to 100 rpm). However, differences reduce at 6% only
for higher TSR values (TSR = 0.16, turning velocity equal to 200 rpm). Results indicate that as the
turning velocity increases, the optimum value of the blade tip exit angle shifts towards smaller values.

Table 3. Torque and power coefficients as a function of the blade orientation angle. All cases use
a 22-blade rotor with U = 2 m s−1 (Re = 1.7 × 104 ).

TSR = 0.08 TSR = 0.16

Blade Tip Exit Angle Cm Cp Cm Cp
20.5◦ 0.798 0.065 0.692 0.112
25.5◦ 0.857 0.070 0.723 0.117
30.5◦ 0.896 0.073 0.730 0.118
35.5◦ 0.846 0.069 0.687 0.112

6. HAWT
Spinning crossflow runners mounted as blades in horizontal-axis turbines work in the regime
0 < α (see Table 2 and Figure 2) so the energy required for spinning them must be externally supplied
(by an electrical motor, for example). The net lift over the runner provides the force that generates
the torque for rotating the blades around the center of the hub (see Figure 1). Here, the aerodynamic
coefficients Cl and Cd , rather than the torque coefficient Cm , determine the performance of the turbine.
The 2D behavior of the 22-blade crossflow runner is investigated with ANSYS-Fluent for incoming
flow velocities U = 2 m s−1 , 5 m s−1 and 10 m s−1 and spinning turning speeds from 0 to 500 rpm at
100 rpm intervals (turning clockwise; that contrasts with Sections 4 and 5 where the runner turned
anticlockwise). As an example of the results found, the N = 300 rpm and U = 2 m s−1 (Re = 8489) case is
discussed. For this case, Figure 18 shows drag and lift coefficients of the 22 blades as a function of time.
The frequency analysis of the signal of Figure 18 reveals a dominant peak at 10 Hz with harmonics
of decreasing amplitude that exhibit a secondary peak at 110 Hz. The latter corresponds to the blade to
blade passing frequency (f = 22N/60 with N = 300 rpm) of the crossflow runner whilst the 10 Hz value
is related to the signal of the oscillating flow past the runner as described next.
Energies 2018, 11, 110 16 of 24
Energies 2018, 11, 110 16 of 24
Energies 2018, 11, 110 16 of 24

Figure 18. 18.

Figure Drag
Drag CdCdand
and lift CCl lcoefficients
coefficients as aasfunction
a function
of timeof for
timethefor the 22(Nblades
22 blades (N =U300
= 300 rpm, =2m rpm,
Figure
U = 2s m
−1, Re −
18. 1 Drag
s = ,8489). C and
Re = 8489).
d lift C l coefficients as a function of time for the 22 blades (N = 300 rpm, U = 2 m
s−1, Re = 8489).
Figure 19 shows the pressure field at 0.04 s time intervals starting at 9.11 s and ending at 9.21 s.
Figure 19 shows the pressure field at 0.04 s time intervals starting at 9.11 s and ending at 9.21 s.
TheFigure
pressure 19 shows
pattern the pressure
clearly field at
indicates the0.04 s timeof
existence intervals
a vortexstarting
shedding at at
9.11thes upper
and ending
left exitat of
9.21
thes.
The pressure pattern clearly indicates the existence of a vortex shedding at the upper left exit of the
Therunner.
pressure A frequency
pattern clearly analysis by means
indicates the of the FastofFourier
existence a vortex Transform
sheddingofatthe thepressure
upper leftsignal
exitatofthe
the
runner.
pointAA
runner. frequency
located
frequency analysis
analysis by
D/2 downstream byfrommeans
means of the
the of the of
center Fast Fourier
rotation
Fast Fourier Transform
andTransform
D/4 above of itof thepressure
confirms
the pressure
a maximum signal
signal at
thethe
peak
at
point located
at frequency
point located D/2D/2 downstream
f =downstream
9.9 Hz, followed from
fromwith the center
the harmonics of rotation
and a secondary
center of rotation and
and D/4 aboveD/4 above
peak at it confirms
the bladeatomaximum
it confirms a maximum
blade passing peak
peak at frequency
frequency
at frequency = 9.9fHz)
f(110 = 9.9
Hz, Hz,Figure
(see
followed followed
with with
20).harmonics
As harmonics
expected,and the and
110 Hz
a secondary a secondary
peak peak
peak isat much
the blade at the
narrower blade
to blade than tothat
passingblade
passing frequency
corresponding to(110
the Hz)
signal (see
of Figure
the vortex20). As
shedding expected,
due to the
the
frequency (110 Hz) (see Figure 20). As expected, the 110 Hz peak is much narrower than that 110
inherent Hz peak
variability is much
of the narrower
latter signal. than
that The Strohual
corresponding numberto theof the
signalvortex of shedding
the vortex is St = fD/U
shedding = 0.31
due that
to for
the
corresponding to the signal of the vortex shedding due to the inherent variability of the latter signal. α = 0.49
inherent is close to the
variability condition
of the latter
TheofStrohual
signal. the
Thefirst shedding
Strohual
number ofmode
number ofinthe
the vortex alternate
vortexvortex
shedding is Stdetected
shedding is St=in
= fD/U = circular
fD/U
0.31 that=forcylinders
0.31 0.49(although
α =that for tofor
α = 0.49
is close theishigher
close to
conditionRe the
number
condition
of the first [23]).
of shedding
the first shedding
mode in alternatemode invortex alternate vortexindetected
detected in circular
circular cylinders cylinders
(although for(although
higher Refor
higher Re [23]).
number number [23]).

Figure 19. Pressure contours at different times for N = 300 rpm (clockwise) and U = 2 m s−1 (Re = 8489).

Figure19.
Figure 19.Pressure
Pressurecontours
contoursatatdifferent
differenttimes
timesfor
forN
N == 300
300 rpm
rpm (clockwise)
(clockwise) and
andUU==22mms s−(Re
−1 = 8489).
1 (Re = 8489).
Energies Energies 11011, 110
2018, 11,2018, 17 of 24 17 of 24
Energies 2018, 11, 110 17 of 24

Figure 20. Frequency spectra of the pressure signal at a point situated at the wake of the crossflow
Figure
Figure runner (seeFrequency
20. text) (N
20. Frequency
spectra
= 300
spectra rpm of theUpressure
and
of the ,signal
= 2 m s−1signal
pressure
at a point situated at the wake of the crossflow
Re = 8489).
at a point situated at the wake of the crossflow
runner (see text) (N = 300 rpm and U = 2 m s−1 , Re = 8489).
runner (see text) (N = 300 rpm and U = 2 m s−1, Re = 8489).
The time evolution of drag and lift coefficients for a single blade of the crossflow runner (Figure
21) clearly
The timehasevolution
a periodicofsignal with
drag and lifta coefficients
frequency equal to thatblade
for a single of theofturning velocity
the crossflow N/60(Figure
runner (=5 Hz 21)
for
The time
the N =has
clearly evolution
300a periodicof
rpm case). drag and
Besides
signal lift
with this coefficients
dominant
a frequency for
peak,
equal a
the of
to that single
frequencyblade
the turning of the
analysis crossflow
(not
velocity shown)
N/60 (=5 Hzrunner
for thea(Figure
reveals
21) clearly
N has
secondary a periodic
= 300 rpmpeak signal
at Besides
case). with
10 Hz correspondinga frequency
this dominant to the the
peak, equal
dominant to
frequency that of
frequency the turning
analysisof(not
theshown) velocity
vortex reveals
shedding N/60 (=5 Hz for
observed
a secondary
the N =peak
in300 atrpm
Figure Hzcase).
1019. Besidestothis
corresponding dominant
the dominant peak, the
frequency of thefrequency analysis
vortex shedding (not shown)
observed reveals a
in Figure 19.
secondary peak at 10 Hz corresponding to the dominant frequency of the vortex shedding observed
in Figure 19.

Figure 21.
Figure 21. Drag
Drag Cd
Cd (a);
(a); and
and lift
lift Cl
Cl (b)
(b) coefficients
coefficients as
as aa function
function of
of time
time for
for one
one blade
blade of
of the
the crossflow
crossflow
runner (N = 300 rpm, U = 2 m s −1
− ,
1 Re =
runner (N = 300 rpm, U = 2 m s , Re = 8489). 8489).

We note that Figure 21 predicts positive and negative values of Cd and Cl depending on the blade
We note that Figure 21 predicts positive and negative values of Cd and Cl depending on the blade
position of the crossflow runner. This behavior can be better observed through the polar graph shown
Figure 21. of
position Drag Cd (a); and
the crossflow lift ClThis
runner. (b) coefficients
behavior canas beabetter
function of time
observed for one
through theblade
polarof the crossflow
graph shown
in Figure 22 that represents the values of Cd and Cl corresponding to one blade (data from 7 s to 7.2
in Figure
runner =22
(Ndiscussthat
300 represents
rpm, U=2m the values of Cd and Cl corresponding to one blade (data from 7 s to 7.2 s).
of sFigure
, Re =228489).
−1
s). We the results with the aid of the velocity vectors and pressure contours at
We discuss the results of Figure 22 with the aid of the velocity vectors and pressure contours at the
the simulated plane at t = 8.334 s (see Figure 23). The geometrical position of this case is similar to
We simulated
note thatplane at t = 8.334
Figure s (see Figure 23). Thenegative
geometrical position
of Cof this Ccase is similar to that
that of the 9.13 s case21in predicts
Figure 19 positive and
since the flow pattern near values
the blades d and
of l depending
the crossflow runneron the blade
does
of the 9.13 s case in Figure 19 since the flow pattern near the blades of the crossflow runner does not
positionnotofsubstantially
the crossflow varyrunner.
in time This behavior
(see Figure 19). can be better observed through the polar graph shown
substantially vary in time (see Figure 19).
in Figure 22 that 23
Figure represents
is used forthe valuesthe
explaining of performance
Cd and Cl corresponding
of one single bladeto of
onetheblade
crossflow(data from
runner as7it s to 7.2
s). We rotates
discuss the results
(Figure of Figure
22). A blade 22 withrunner
of the crossflow the aid of the an
generates velocity vectors
intense lift and pressure
in the region 180◦ < θ <contours
240◦ at
the simulated
(Figure 22)plane at t = 8.334 to
that corresponds s (see Figure
the zone 23).
where itsThe geometrical
orientation is moreposition
favorableof forthis case isthis
producing similar to
that of aerodynamic
the 9.13 s case force
in (see Figure
Figure 19 23).
since The
theexistence of this zone
flow pattern near was already of
the blades observed in the analysis
the crossflow runner does
of VAWT (anticlockwise rotation;
not substantially vary in time (see Figure 19). Figure 16). The main differences are that, under clockwise rotation,
the inner vortex: (1) is less intense; and (2) has a center closer to the exit of the lower right upstream
blades (verified by comparing cases with equal U and N but with different sense of rotation).
In comparison with the regimes studied in the VAWT section, the forced clockwise rotation
modifies the triangles of velocities at both blade inlet and outlet points, leading to greater outflow
velocities at the extrados of the lower left downstream blades. This implies higher pressure differences
Figure 22. Drag Cd (a); and lift Cl (b) coefficients for one single blade of the crossflow runner (θ = 0°
corresponds to the blade location in the upper position in Figure 4; N = 300 rpm, U = 2 m s−1, Re = 8489).
 Figure 21. Drag Cd (a); and lift Cl (b) coefficients as a function of time for one blade of the crossflow
runner (N = 300 rpm, U = 2 m s−1, Re = 8489).

Energies
We note 11, 110
2018,that Figure 21 predicts positive and negative values of Cd and Cl depending on18the
of 24
blade
position of the crossflow runner. This behavior can be better observed through the polar graph shown
in Figure
between 22 the
thatupper
represents
and thethe values
lower of of
sides Cdthe
andblades
Cl corresponding
and therefore to one blade
higher (dataIn
lift values. from 7 s to 7.2
addition,
s). We
the discuss
pattern ofthetheresults of Figure
static pressure 22 with the
surrounding theaid of the
entire velocity
crossflow vectors
runner seenand pressure
in Figure contours at
23 resembles
the that
simulated planerotating
of clockwise at t = 8.334 s (see
cylinders Figure
with a net 23). The differences
pressure geometrical positionaof
producing this case
positive is similar to
lift (Magnus
thatforce).
of theThe
9.13position of the
s case in blade19with
Figure thethe
since greatest
flow contribution
pattern neartothe
drag is 60◦ <
blades < 150
ofθthe ◦ that corresponds
crossflow runner does
not to a zone with almost
substantially vary inatime
perpendicular
(see Figure hit19).
with the incoming flow (Figure 23).

Energies 2018, 11, 110 18 of 24

Figure 23 is used for explaining the performance of one single blade of the crossflow runner as
it rotates (Figure 22). A blade of the crossflow runner generates an intense lift in the region 180° < θ
< 240° (Figure 22) that corresponds to the zone where its orientation is more favorable for producing
this aerodynamic force (see Figure 23). The existence of this zone was already observed in the analysis
of VAWT (anticlockwise rotation; Figure 16). The main differences are that, under clockwise rotation,
the inner
Figure 22.vortex:
Drag
DragC(1) is less
d (a); and
Cd (a);
intense;
andlift
liftCC
l (b)and (2) has afor
coefficients center
one closer
single to the exit
blade of crossflow
of the the lowerrunner
right upstream
(θ◦ = 0°
Figure 22. l (b) coefficients for one single blade of the crossflow runner (θ = 0
blades (verified
corresponds
corresponds by
to the
to thecomparing
blade
bladelocation cases
locationin with
inthe equal
the upper U and N but
positionininFigure
upper position Figurewith different
4; =N300
4; N = 300 rpm,
rpm, sense
U =U 2m of rotation).
= 2s m, sRe,=Re
− 1 −1 = 8489).
8489).

Figure 23.Velocity
Figure23. Velocityvectors
vectors(a);
(a);and
andpressure
pressurecontours
contours(b) attt== 8.334
(b)at 8.334 s for N == 300
300 rpm,
rpm, U mss−1−, 1Re
U ==22m ,
Re= = 8489).
8489).

OnIn the
comparison
other hand, withthethe
liftregimes studied
coefficient Cl forinthetheentire
VAWT section,
runner the aforced
shows steadyclockwise rotation
increase with N
modifies the triangles of velocities at both blade inlet and outlet points, leading
or, equivalently, with α (see Figure 24, in which, for completeness, data corresponding to Section 5 to greater outflow
velocities
are included). at For
the comparison
extrados ofpurposes,
the lower left downstream
Cl values corresponding blades. This implies
to rotating cylinders higher pressure
are taken into
differences
account. Thebetween the value
theoretical upper foranda the lower
plain sidescylinder
rotating of the blades
appliesandthe
therefore higher lift values.
Kutta–Joukowski theoremIn
addition,
(see, the pattern
e.g., [37]), being of the static pressure surrounding the entire crossflow runner seen in Figure 23
resembles that of clockwise rotating cylinders withΓ a net pressure differences producing a positive
Cl = , (7)
lift (Magnus force). The position of the blade with U the greatest contribution to drag is 60° < θ < 150°
that corresponds
where Г = 2πrω is to thea circulation,
zone with almost
with rathe
perpendicular
radius of thehit with the
cylinder incoming
and flow (Figure
ω its angular 23).Thus,
velocity.
On the
the dotted other
line hand, 24
in Figure thehas
liftacoefficient
slope equalCl to
for2πthe entirethe
whilst runner
shadedshows a steady
region increase
includes with N
measured andor,
equivalently,
simulated datawith
of Cαl for
(seecylinders
Figure 24, inin which,
the rangefor of completeness,
6 < U < 22 m data s−1 and
corresponding
0 < N < 3000 to rpm
Section 5 are
[38,39].
Inincluded).
Figure 24,For thecomparison purposes,
crossflow runner Cl values
provides highercorresponding
lift coefficientstothan
rotating
thosecylinders
of cylindersare (with
takenandinto
account.spirals)
without The theoretical
at very low value
freefor a plain
stream rotating cylinder applies the Kutta–Joukowski theorem (see,
velocities.
e.g., [37]), being
𝛤
𝐶𝑙 = , (7)
𝑈
where Г = 2πrω is the circulation, with r the radius of the cylinder and ω its angular velocity. Thus,
the dotted line in Figure 24 has a slope equal to 2π whilst the shaded region includes measured and
simulated data of Cl for cylinders in the range of 6 < U < 22 m s−1 and 0 < N < 3000 rpm [38,39]. In
Figure 24, the crossflow runner provides higher lift coefficients than those of cylinders (with and
Energies
Energies 2018, 11, 110 19
19 of
of 24
24

Energies 2018, 11, 110 19 of 24

Figure 24. Lift coefficient Cl as a function of α for a crossflow runner with 22 blades. Negative values
Figure 24. Lift coefficient Cl as a function of α for a crossflow runner with 22 blades. Negative values
of α correspond to an anticlockwise turning (analyzed in Section 5). Data for a plain cylinder from
of α correspond to an anticlockwise turning (analyzed in Section 5). Data for a plain cylinder from
Refs. [38,39] and for a cylinder with spirals from Ref. [22].
Refs. [38,39] and for a cylinder with spirals from Ref. [22].

However, the relevant parameter for determining the performance of lift-based wind turbines is
the drag Cd to lift Cl ratio rather than the lift coefficient Cl only. For this type of turbine, results indicate
a dependence of Cd /Cl on 10α valid for U ≤ 10 m s−1 (see Figure 25). Simulations are limited to α
values less than 0.8 since greater figures are achieved with unrealistic very high spinning velocities of
the crossflow runner. Values of Cd /Cl for crossflow runners are as low as those recently obtained in
Figure 24. Lift
state-of-the-art coefficientairfoils
circulating Cl as a (although
function of the
α forlatter
a crossflow
can berunner
appliedwith
to 22 blades.range
a larger Negative values
of α). Overall,
of α correspond to an anticlockwise turning (analyzed in Section 5). Data for a plain cylinder
the Cd /Cl ratio for crossflow turbines is on the order of 10 times smaller than that of circulating from
Refs. at
cylinders [38,39] and for
the same a cylinder
value of α. with spirals from Ref. [22].

Figure 25. Drag to lift ratio for crossflow runners with 22 blades as a function of α and comparison
with rotating cylinders and circulating airfoils.

According to the flow process diagram (Figure 2), an analytical approximation is applied for
estimating the Cp value of crossflow HAWT. At spinning ratios α higher than 0.3, the flow at the wake
of a crossflow runner clearly resembles that of a spinning cylinder at similar Re (see Figure 26; [39]).
Therefore, the Cp equation is based on the analytical formulation developed by Sedaghat [15], who
studied the performance of Magnus wind turbines (HAWT with spinning cylinders).

Figure 25.
Figure 25. Drag
Drag to
to lift
lift ratio
ratio for
for crossflow
crossflow runners
runners with
with 22
22 blades
blades as
as aa function
function of
of αα and
and comparison
comparison
with rotating
with rotatingcylinders
cylindersand andcirculating
circulatingairfoils.
airfoils.

According to
According to the
the flow
flow process
process diagram
diagram (Figure
(Figure 2),
2), an
an analytical
analytical approximation
approximation is is applied
applied for
for
estimatingthe
estimating the CCp value
value ofof crossflow HAWT.At
crossflow HAWT. At spinning
spinningratios
ratios ααhigher
higherthan
than0.3,
0.3, the
the flow
flow atat the
the wake
wake
p
of a crossflow runner clearly resembles that of a spinning cylinder at similar Re (see Figure 26; [39]).
of a crossflow runner clearly resembles that of a spinning cylinder at similar Re (see Figure 26; [39]).
Therefore, the C p equation is based on the analytical formulation developed by Sedaghat [15], who
Therefore, the Cp equation is based on the analytical formulation developed by Sedaghat [15], who
studied the
studied the performance
performance of of Magnus
Magnus wind
wind turbines
turbines(HAWT
(HAWTwith withspinning
spinningcylinders).
cylinders).

Figure 26. Streamlines: (a) at t = 8.334 s for N = 300 rpm (α = 0.49); and (b) at t = 2.141 s for N = 400 rpm
(α = 0.65) (clockwise rotation). U = 2 m s−1, Re = 8489.
 According to the flow process diagram (Figure 2), an analytical approximation is applied for
estimating the Cp value of crossflow HAWT. At spinning ratios α higher than 0.3, the flow at the wake
of a crossflow runner clearly resembles that of a spinning cylinder at similar Re (see Figure 26; [39]).
Therefore,
Energies 2018,the Cp equation is based on the analytical formulation developed by Sedaghat [15],20who
11, 110 of 24
studied the performance of Magnus wind turbines (HAWT with spinning cylinders).

Figure 26.
Figure 26. Streamlines:
Streamlines: (a)
(a) at
at tt == 8.334
8.334 ss for
for N N == 300
300 rpm
rpm (α
(α == 0.49);
0.49); and
and (b)
(b) at
at tt == 2.141
2.141 ss for
for N
N == 400
400 rpm
rpm
(α = 0.65) (clockwise rotation). U = 2 m s −1
− ,
1 Re
(α = 0.65) (clockwise rotation). U = 2 m s , Re = 8489. = 8489.

The analytical approximation applies the blade element-momentum theory in which the blade
is divided into elements of infinitesimal thickness dr along the radial direction. The local equations
of a blade element located at an arbitrary distance r take into account the value of the relative wind
incidence angle, which changes with r. The power coefficient Cp is obtained after integrating a local
power coefficient term from hub to tip radius (see [15] for details), being

Z 1
" #
a ( 1 − a ) (1 − ε tanβ)
C p = 8a(1 − a)λ R µ2 + 2
dµ, (8)
µ0 λR (ε + tanβ)

where
r Ωr C λr (1 + a0 ) − α/2
µ= ; λr = ; ε = d ; tanβ = , (9)
R U Cl 1−a
with r the radial distance from the center of the hub (of radius r0 ) that rotates at Ω, and R the radius of
the turbine. In Equation (8), the axial flow induction factor a = 1/3 and the angular induction factor
a’ follows
a (1 − a ) α
a0 = + (10)
λ2r 2λr
Equation (8) is an integration from r = r0 (µ = µ0 ) to r = R (µ = 1) of the local contribution to Cp by
taking into account the changes along the radial direction of the relative wind velocity with respect to
the crossflow runner. Once µ0 is kept fixed, Equation (8) depends only on the global spin ratio of the
crossflow runner α and on the TSR (= λ R ) of the turbine. The drag to lift ratio ε used in Equation (8) is
obtained after adjusting the data of Figure 25, giving

Cd
ε= = 0.863e−2.292αw , (11)
Cl

with an R-square equal to 0.981.

In Equation (11), α is substituted by the local spin ratio αW defined as

ωD/2
αW = , (12)
W
with W the relative velocity of the air with respect to the blade. Thus, Equations (8) and (11) take into
account that, for fixed values of TSR and global spin ratio α, the resultant velocity over the crossflow
runner varies along the radial distance r from the center of the hub, being
 1/2
W = U (1 − a) 1 + tan2 β , (13)
Energies 2018, 11, 110 21 of 24

and thus,
1
αW = α
1/2 . (14)
(1 − a) 1 + tan2 β
For µ0 = 0.05, the power coefficient as a function of α and TSR is shown in Figure 27 where
Equation (11) is assumed to be valid up to αW ≈ 2. From Figure 25, this means a drag to lift ratio for
a crossflow runner similar to that from circulating airfoils. Results indicate that the optimum TSR for
HAWT is ≈0.8
Energies 2018, for any spin ratio. At α = 1, Cp = 0.41, which represents an increase of 18% with respect
11, 110 21 of 24
to the performance of Magnus type HAWT (Table 1). For example, we estimate that a small HAWT
where
with the last equality
crossflow indiameter
runners of EquationD(15) considers
= 0.062 m andthe definition
length ≈ 1 mof the torque
should coefficient
spin at s−1 and
in Equation
ω = 64.5 rad
(5) and,
rotate as discussed
with the hub at Ωit=uses
respect to previously, 1.6 rad s−1 for having
the resultant relative
a Cvelocity
p = 0.41 over
whentheU=blade −1 .
2 m sW.

Figure 27.
Figure 27. CCpp (a); and CCpnet
pnet (b)
(b) (for
(for turbine
turbine solidity
solidity σσ == 0.04)
0.04) of
of a HAWT with crossflow
crossflow runners
runners as
as
blades as
blades as aa function
function ofof TSR
TSR (=ΩR/U)
(=ΩR/U) and and spin
spin ratio
ratio αα (=ωD/(2U)).
(=ωD/(2U)).

Fromrequired
The Equations (2), (9),power
external (13) and (15),spinning
Pr for we definethethe requiredrunners
crossflow in HAWT 𝐶
power coefficient as
is𝑝𝑟estimated by
assuming that the required torque𝑃balances the 1
resistant one generated by the flow over the entire
𝑟
runner. Thus, 𝐶𝑝𝑟 = = 𝜎𝛼(1 − 𝑎)2 ∫ 𝐶𝑚 (1 + 𝑡𝑎𝑛2 𝛽)𝑑𝜇 , (16)
𝑃𝑈 Z Z R𝜇0 2
1 D
Pr = ω dM = ω Cm ρW 2 dr, (15)
where PU is the power of the incoming flow in a surface
r0 area
2 swept
2 by the turbine and σ the turbine
solidity defined as
where the last equality in Equation (15) considers the definition of the torque coefficient in Equation (5)
1
and, as discussed previously, it uses the resultant 𝐵𝐷𝑅 over the blade W.
relative velocity
𝑃𝑈 = 𝜌𝑈 3 𝜋𝑅2 ; 𝜎 = , (17)
From Equations (2), (9), (13) and (15), we 𝜋𝑅2 power coefficient C pr as
2 define the required
with B the number of crossflow runners (i.e., of blades).
Z 1
Pr  
From the above, the available
C pr = net=power
σα(1 − a)net2 = P C
is P − mPr,1from 2
which
+ tan the net power coefficient (16)
β dµ, Cpnet
PU µ0
is defined as
where PU is the power of the incoming flow in𝑃a−surface 𝑃𝑟 area swept by the turbine and σ the turbine
𝐶𝑝𝑛𝑒𝑡 = = 𝐶𝑝 − 𝐶𝑝𝑟 (18)
solidity defined as 𝑃𝑈
1 3 2 BDR
Equation (18) depends only onPthe U =tip ρU
speed ; σ =TSR, the
πRratio , spin ratio of the runners α and(17) the
2 πR2
turbine solidity σ. The number of blades B in the blade element-momentum theory developed by
with B the number of crossflow runners (i.e., of blades).
Sedaghat [15] does not directly affect the global power coefficient, although it is taken into account
From the above, the available net power is Pnet = P − Pr , from which the net power coefficient
in the calculation of the global torque. However, B indirectly affects the value of the net power
Cpnet is defined as
coefficient since it modifies the turbine solidity value (an increase in B reduces Cpnet). Nevertheless,
P − Pr
the effect of varying the number of bladesC pnet = = C p − C prof the HAWT would require complex
on the performance (18)
PU
3D CFD simulations.
Equation (18) in
The Cm value depends only
Equation onisthe
(16) tip speed
estimated by ratio TSR,
fitting the the spin ratio
simulated of the
torque runners
data andlocal
againstα the the
turbine solidity
for theσ.UThe number blades𝐶B in
of being the blade 2 element-momentum theory developed by
spin ratio =2m s−1 case, 𝑚 = 3.312 𝛼𝑊 + 2.387𝛼𝑊 + 0.925, with an R-square equal
to 0.995. We do not expect substantial changes of Cm for other velocities, as inferred from Figure 17.
Results of Cpnet for a solidity σ = 0.04 are also shown in Figure 27. In comparison with the Cp value, the
available net power drastically reduces and makes unfeasible the applications of HAWT at high α
due to the high demand of power for spinning the runners.
Energies 2018, 11, 110 22 of 24

Sedaghat [15] does not directly affect the global power coefficient, although it is taken into account
in the calculation of the global torque. However, B indirectly affects the value of the net power
coefficient since it modifies the turbine solidity value (an increase in B reduces Cpnet ). Nevertheless,
the effect of varying the number of blades on the performance of the HAWT would require complex
3D CFD simulations.
The Cm value in Equation (16) is estimated by fitting the simulated torque data against the local
spin ratio for the U = 2 m s−1 case, being Cm = 3.312αW 2 + 2.387α + 0.925, with an R-square equal
W
to 0.995. We do not expect substantial changes of Cm for other velocities, as inferred from Figure 17.
Results of Cpnet for a solidity σ = 0.04 are also shown in Figure 27. In comparison with the Cp value,
the available net power drastically reduces and makes unfeasible the applications of HAWT at high α
due to the high demand of power for spinning the runners.

7. Conclusions
The potential use of crossflow runners as a single rotor in VAWT and as blades in HAWT has been
numerically investigated. The 2D CFD model (ANSYS-Fluent) has been validated with experimental
data obtained in a wind tunnel under runaway conditions. All cases have been simulated using the
same domain as the one used in the experimental set up, so small variations in the results may occur
when working in an open environment.
The working mode of crossflow runners in VAWT has a sense of rotation opposed to those runners
used in HAWT. In VAWT, the crossflow runner rotates in the same direction than in the free case
(autorotation or runaway regime). All working modes show that crossflow runners combine both
drag-driven (upstream blades) and lift-driven (downstream blades) mechanisms for generating forces
and torque, being the lift effect the most important factor in all cases. An inner vortex within the
crossflow runner is formed, being more relevant when working in VAWT regimes than in HAWT ones.
The power coefficient Cp in VAWT is maximum at very low values of TSR (≈0.3), varying from
0.19 (our work) to 0.45 [9]. The discrepancy comes from working with different blade shapes, aspect
ratios and Reynolds numbers (2.1 × 104 here and 1.3 × 105 in [9]). In general, high solidity runners are
preferred for working in VAWT.
Crossflow runners used as blades in HAWT require an external (non-wind) power supply for
spinning the runners at a fixed turning velocity. Drag, lift and torque coefficients over the whole runner
for a 2D simulation show an oscillatory behavior with a dominant frequency that does not coincide
neither with the runner spin frequency nor with the runner’s blade-to-blade one. This dominant
frequency has its origin in the non-alternating vortex shedding detected downstream the runner and,
because of its amplitude, it should be taken into account for avoiding resonances when designing the
structural system of the turbine. This effect is not clearly observed when working in the VAWT mode.
The drag to lift ratio of runners for HAWT is comparable to that of circulating airfoils and
~10 times lower than that for cylinders at the same spinning ratios α. This leads to values of Cp of
a HAWT obtained by applying an analytical approximation 18% higher than those expected in Magnus
type wind turbines (i.e., Cp = 0.41 for TSR = 0.7 and α = 1). However, the available net power (extracted
minus required for spinning the cylinders) is much lower, with a maximum net power coefficient
Cpnet = 0.14 for TSR = 0.9 and α = 0.7. Thus, for HAWT applications, it is essential to design crossflow
runners with aerodynamic blades to have very low drag to lift ratios and, at the same time, low
resistant torques.

Acknowledgments: This work has been partially funded by the University of Girona under grant MPCUdG2016-4.
Author Contributions: Eduard Massaguer and Albert Massaguer conducted the experimental study, including the
set up of the system. Toni Pujol and Lino Montoro conducted the detailed modeling for all cases. Martí Comamala
and Toni Pujol carried out the analyses of the results. Toni Pujol and Albert Massaguer wrote the paper.
Conflicts of Interest: The authors declare no conflict of interest. The founding sponsors had no role in the design
of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the
decision to publish the results.
Energies 2018, 11, 110 23 of 24

References
1. Ackermann, T.; Söder, L. Wind energy technology and current status: A review. Renew. Sustain. Energy Rev.
2000, 4, 315–374. [CrossRef]
2. Dixon, S.L.; Hall, C.A. Fluid Mechanics and Thermodynamics of Turbomachinery, 6th ed.; Butterworth-Heinemann:
Burlington, MA, USA, 2010.
3. Cooney, C.; Byrne, R.; Lyons, W.; O’Rourke, F. Performance characterization of a commercial-scale wind
turbine operating in an urban environment, using real data. Energy Sustain. Dev. 2017, 36, 44–54. [CrossRef]
4. Li, J.; Yu, X. Analyses of the Extensible Blade in Improving Wind Energy Production at Sites with Low-Class
Wind Resource. Energies 2017, 10, 1295. [CrossRef]
5. Tummala, A.; Velamati, R.K.; Sinha, D.K.; Indraja, V.; Krishna, V.H. A review on small scale wind turbines.
Renew. Sustain. Energy Rev. 2016, 56, 1351–1371. [CrossRef]
6. Yang, A.-S.; Su, Y.-M.; Wen, C.-Y.; Juan, Y.-H.; Wang, W.-S.; Cheng, C.-H. Estimation of wind power generation
in dense urban area. Appl. Energy 2016, 171, 213–230. [CrossRef]
7. Chong, W.-T.; Fazlizan, A.; Poh, S.C.; Pan, K.C.; Hew, W.P.; Hsiao, F.B. The design, simulation and testing of
an urban vertical axis wind turbine with the omni-direction-guide-vane. Appl. Energy 2013, 112, 601–609.
[CrossRef]
8. Santhakumar, S.; Palanivel, I.; Venkatasubramanian, K. A study on the rotational behaviour of a Savonius
Wind turbine in low rise highways during different monsoons. Energy Sustain. Dev. 2017, 40, 1–10. [CrossRef]
9. Dragomirescu, A. Performance assessment of a small wind turbine with crossflow runner by numerical
simulations. Renew. Energy 2011, 36, 957–965. [CrossRef]
10. Al-Maaitah, A.A. The design of the Banki wind turbine and its testing in real wind conditions. Renew. Energy
1993, 3, 781–786. [CrossRef]
11. Gupta, R.; Biswas, A.; Sharma, K.K. Comparative study of a three-bucket Savonius rotor with a combined
three-bucket Savonius-three-bladed Darrieus rotor. Renew. Energy 2008, 33, 1974–1981. [CrossRef]
12. Tian, W.; Song, B.; VanZwieten, J.H.; Pyakurel, P. Computational Fluid Dynamics Prediction of a Modified
Savonius Wind Turbine with Novel Blade Shapes. Energies 2015, 8, 7915–7929. [CrossRef]
13. Rezaeiha, A.; Kalkman, I.; Blocken, B. Effect of pitch angle on power performance and aerodynamics of
a vertical axis wind turbine. Appl. Energy 2017, 197, 132–150. [CrossRef]
14. Ying, P.; Chen, Y.K.; Xu, Y.G.; Tian, Y. Computational and experimental investigations of an omni-flow wind
turbine. Appl. Energy 2015, 146, 74–83. [CrossRef]
15. Sedaghat, A. Magnus type wind turbines: Prospectus and challenges in design and modelling. Renew. Energy
2014, 62, 619–628. [CrossRef]
16. Hirahara, H.; Hozzain, M.Z.; Kawahashi, M.; Nonomura, Y. Testing basic performance of a very small wind
turbine designed for multi-purposes. Renew. Energy 2005, 30, 1279–1297. [CrossRef]
17. Bychkov, N.M.; Dovgal, A.V.; Kozlov, V.V. Magnus wind turbines as an alternative to the blade ones. J. Phys.
Conf. Ser. 2007, 75, 1–7. [CrossRef]
18. Borgh, J. Magnus Effect: An Overview of Its Past and Future Practical Applications; Technical Report; Defense
Technical Information Center: Fort Belvoir, VA, USA, 1986; Volumes 1–2.
19. Murakami, N.; Ito, J. Magnus Type Wind Power Generator. U.S. Patent 7,504,740 B2, 17 March 2009.
20. Giudice, F.; La Rosa, G. Design, prototyping and experimental testing of a chiral blade system for
hydroelectric microgeneration. Mech. Mach. Theory 2009, 44, 1463–1484. [CrossRef]
21. Gono, R.; Rusek, S.; Hrabcik, M. Wind turbine cylinders with spiral fins. In Proceedings of the 8th EEEIC
International Conference on Environment and Electrical Engineering, Karpacz, Poland, 10–13 May 2009;
pp. 41–45.
22. Mecaro Co. Japan. Available online: www.mecaro.jp/eng/ (accessed on 3 October 2016).
23. Kazemi, S.A.; Nili-Ahmadabadi, M.; Sedaghat, A.; Saghafian, M. Aerodynamic performance of a circulating
airfoil section for Magnus systems via numerical simulation and flow visualization. Energy 2016, 104, 1–15.
[CrossRef]
24. Mehta, R.D.; Bradshaw, P. Design rules for small low speed wind tunnels. Aeronaut. J. R. Aeronaut. Soc. 1979,
443–449.
25. Etemaddar, M.; Hansen, M.O.L.; Moan, T. Wind turbine aerodynamic response under atmospheric icing
conditions. Wind Energy 2014, 17, 241–265. [CrossRef]
Energies 2018, 11, 110 24 of 24

26. Seifert, J. A review of the Magnus effect in aeronautics. Prog. Aerosp. Sci. 2012, 55, 17–45. [CrossRef]
27. Xisto, C.M.; Páscoa, J.C.; Leger, J.A.; Trancossi, M. Wind energy production using an optimized variable
pitch vertical axis rotor. In Proceedings of the ASME 2014 International Mechanical Engineering Congress &
Exposition, Montreal, QC, Canada, 14–20 November 2014. IMECE2014-38966.
28. Chen, G.; Xiong, Q.; Morris, P.J.; Paterson, E.G.; Sergeev, A.; Wang, Y.-C. OpenFOAM for Computational
Fluid Dynamics. Not. AMS 2014, 61, 354–363. [CrossRef]
29. Aramideh, S.; Xiong, Q.; Kong, S.-C.; Brown, R.C. Numerical simulation of biomass fast pyrolysis in an auger
reactor. Fuel 2015, 156, 234–242. [CrossRef]
30. Fagbenro, K.A.; Mohamed, M.A.; Wood, D.H. Computational modeling of the aerodynamics of windmill
blades at high solidity. Energy Sustain. Dev. 2014, 22, 12–20. [CrossRef]
31. Hwang, I.S.; Kang, W.; Kim, S.J. High altitude cycloidal wind turbine system design. Procedia Eng. 2013, 67,
78–84. [CrossRef]
32. Celik, I.B.; Ghia, U.; Roache, P.J.; Freitas, C.J.; Coleman, H.; Raad, P.E. Procedure for estimation and reporting
of uncertainty due to discretization in CFD applications. J. Fluids Eng. 2008, 130, 1–4. [CrossRef]
33. Mittal, S.; Kumar, B. Flow past a rotating cylinder. J. Fluid Mech. 2003, 476, 303–334. [CrossRef]
34. Brusca, S.; Lanzafame, R.; Messina, M. Design of a vertical-axis wind turbine: How the aspect ratio affects
the turbine’s performance. Int. J. Energy Environ. Eng. 2014, 5, 333–340. [CrossRef]
35. Kumar, A.; Saini, R.P. Performance analysis of a Savonius hydrokinetic turbine having twisted blades.
Renew. Energy 2017, 108, 502–522. [CrossRef]
36. Scheurich, F.; Fletcher, T.M.; Brown, R.E. Effect of blade geometry on the aerodynamic loads produced by
vertical-axis wind turbines. Proc. Inst. Mech. Eng. Part A J. Power Energy 2011, 225, 327–341. [CrossRef]
37. White, F.M. Fluid Mechanics, 7th ed.; Mc Graw-Hill: New York, NY, USA, 2011.
38. Tokumaru, P.T.; Dimotakis, P.E. The lift of a cylinder executing rotary motions in a uniform flow. J. Fluid Mech.
1993, 255, 1–10. [CrossRef]
39. Sengupta, T.K.; Kasliwal, A.; De, S.; Nair, M. Temporal flow instability for Magnus-Robins effect at high
rotation rates. J. Fluid Struct. 2003, 17, 941–953. [CrossRef]

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).