Energy 66 (2014) 689e698

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Proposal for an innovative chord distribution in the Troposkien

vertical axis wind turbine concept
Gabriele Bedon, Marco Raciti Castelli, Ernesto Benini*
Department of Industrial Engineering of the University of Padua, Via Venezia 1, 35131 Padua, Italy

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 25 June 2013
Received in revised form
24 December 2013
Accepted 2 January 2014
Available online 28 January 2014

An innovative design for the Troposkien concept is introduced by means of an advanced chord distribution, computed using the WOMBAT (Weatherly Optimization Method for Blades of Air Turbines) algorithm for the performance optimization of vertical axis wind turbines. Five rotor blade architectures,
characterized by a constant value of the thickness-to-chord ratio and a varying chord length along the
blade span, are evaluated. The optimization process is conducted with respect to the power coefcient
for a target wind speed of 9 m/s, obtaining a consistent improvement of rotor performance with respect
to the baseline blade conguration.
2014 Elsevier Ltd. All rights reserved.

Keywords:
Darrieus VAWT (vertical axis wind turbine)
WOMBAT (Weatherly Optimization Method
for Blades of Air Turbines) algorithm
Variable chord distribution
Aerodynamic optimization

1. Introduction
In the near future, VAWTs (vertical axis wind turbines) are going
to play an important role in the energy production planning. In fact,
their typical characteristics of small size and lightweight allow their
installation also in the urban context, on the building roofs and
yards. In this scenario, the Darrieus VAWT represents one the most
promising architecture [1], thanks to its high aerodynamic
performance.
The traditional design of the Darrieus rotor involves a constant
distribution of thickness and chord along the whole blade span. In
the present analysis, the optimal chord distribution along the blade
length is searched for increasing rotor aerodynamic performance.
The Troposkien design procedure usually involves two steps:

the rst step is conducted by considering a BE-M (Blade
Element-Momentum) or Vortex based algorithm, thanks to their
rapidity to provide a reliable estimation of the rotor performance for a given parameter set;

the second step involves the turbine simulation by means of CFD
(Computational Fluid Dynamics) codes, which provide a very

* Corresponding author.
E-mail addresses: gabriele.bedon@dii.unipd.it (G. Bedon), marco.raciticastelli@
unipd.it (M. Raciti Castelli), ernesto.benini@unipd.it (E. Benini).
0360-5442/$ e see front matter 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.energy.2014.01.004

deep knowledge of the rotor aerodynamics, allowing a more

advanced power characterization.
Simulation tools based on the BE-M Theory are widely adopted
and have been proposed and improved by many authors [2e6]. The
main issue regarding these algorithms is their necessity to use
airfoil aerodynamic coefcients for angles of attack over 360 [7e9]
and low airfoil Reynolds numbers, consistently smaller than those
generally considered for aeronautical use [10].
Some aerodynamic databases for a limited number of proles
are available in literature: NACA 0012, SG6043, SD7062 and DU06W-200 polars, characterized by an angle of attack ranging up to
360 and a Reynolds number ranging from 65,000 to 150,000, were
provided by Worasinchai et al. [11]. Sheldahl and Klimas [12] tested
NACA 0009, NACA 0012 and NACA 0015 proles, considering angles
of attacks ranging from 0 to 180 and a limited range of Reynolds
numbers (from 350,000 to 700,000), successively extending the
results to a wider range of Reynolds numbers and also to different
airfoil geometries by means of numerical methods.
In the present work, the WOMBAT (Weatherly Optimization
Method for Blades of Air Turbines) design algorithm developed by
Bedon et al. [13] is adopted. This algorithm includes a rotor aerodynamic performance prediction tool based on the BE-M code
developed by Raciti Castelli et al. [14]. The simulation is based on
the aerodynamic databases provided by Sheldahl and Klimas [12]
and allows also the estimation of the aerodynamic coefcients of

690

G. Bedon et al. / Energy 66 (2014) 689e698

different NACA airfoils by means of the interpolation of such polars.

Nevertheless, for the present calculations, since the prole thickness/chord ratio is maintained constant, this functionality is not
exploited.
The design procedure is conducted optimizing a given baseline
geometry. Nowadays a great number of optimization software
based on genetic algorithms is available on commercial platforms,
allowing the solution of different problems: applications such as
uid dynamics, structural and modal analyses, as well as nancial
evaluations represent some examples. The best aerodynamic
conguration for the rotor blade can therefore be found using a
proper evolutive algorithm, by previously setting the chosen
objective functions. An optimum aerodynamic design model for a
wind turbine blade was presented by Bourguet et al. [15], who
adopted a multi-criteria optimization algorithm coupled to a CFD
simulation code, in order to gain an increase of the nominal power
production and the reduction of the blade weight, suggesting the
adoption of the NACA 0025 prole as the best option. Carrigan et al.
[16] considered a baseline rotor architecture based on a NACA 0015
airfoil and increased the turbine efciency by 6% by means of a
differential evolutionary algorithm based on a CFD simulation code.
The WOMBAT algorithm adopts the multi-objective genetic algorithm based on work of Deb [17] and implemented in the commercial software Matlab [18]. The reliability of such algorithm was
proven by Bedon et al. [19] by optimizing the 2-m Sandia turbine
tested by Sheldahl [20], imposing a constant evolution of both
chord and a thickness along the rotor blade span [13], achieving an
increase in performance up to 6%.
2. The case study
The blades of a Darrieus VAWT are traditionally manufactured
extruding a single prole over the whole blade length, eventually
bent to form a SCS (straight-circular-straight) shape, a practical
approximation for the Troposkien architecture [21].
The 2-m Sandia turbine tested by Sheldahl [20] is here considered as a case study. The rotor is 2 m high, the nominal radius is
0.98 m and the blade shape is SCS. The main geometrical specications of the turbine are reported in Table 1. A picture of the wind
turbine installation in the Sandia site test is reported in Fig. 1.
This type of turbine is widely analyzed in literature, both from
the numerical and from the experimental point of view. The
WOMBAT algorithm has been preliminary tested using such a rotor
architecture [19]: the validation process showed a quite high
sensitivity of the numerical results to the selected dynamic stall
model, especially for lower wind speeds. As a result, the use of the
GormonteStrickland model [22] determined a better prediction
with respect to the experimental data from Sheldahl [20], as can be
seen in Fig. 2. For further information about the WOMBAT optimization algorithm and the validation of its performance prediction code, see Refs. [13,19].
Due to the inherent characteristics of the BEM approach (bidimensional simulation and streamtube independency), the prediction reliability is assumed not to be sensibly compromised by a

Table 1
Main geometrical features of the Troposkien rotor tested by Sheldahl [20],
here assumed as baseline conguration for the optimization process.
H
R
N
Blade prole
Blade shape
c

2m
0.98 m
3
NACA 0012
Straight-circular-straight (SCS)
58.77 mm

Fig. 1. Sandia 2 m Darrieus wind turbine installation (from [20]).

chord variation along the blade span. In order to conrm the consistency of this assumption, an additional validation is shown in
Fig. 3, where WOMBAT predictions are compared to experimental
data from the Sandia 34-m wind turbine [23], characterized by a
variable chord. The aerodynamic coefcients for the SAND 0018/50
prole are retrieved from Ref. [24].
3. Optimization methodology
In the rst optimization campaign, the thickness/chord ratio,
which denes the shape of the prole, is considered uniform and
equal to the value of the Sandia turbine (0.12), considered as a
baseline conguration. In order to have a fair comparison, the same
air density (1 kg/m3 [25]) registered at the Sandia test site is here
considered.
Due to its numerical nature, the algorithm can be adapted to
accept a non uniform chord distribution. Anyway, experimental
data for small turbines with variable chord are not available in
literature: the presented numerical results should therefore be
conrmed by further experimental activity.
As described in Ref. [13], the WOMBAT algorithm is structured as
a loop which involves a genetic algorithm, a BE-M simulation code

G. Bedon et al. / Energy 66 (2014) 689e698

691

Fig. 2. Comparison between experimental data and BE-M predictions (based on the
aerodynamic database provided by Ref. [12]) without and with different dynamic stall
models: Gormont-Strickland (GS) and Gormont-Berg (GB). Rotor angular speed is
460 rpm (from [19]).

and an aerodynamic coefcient predictor. This last feature is not

used in the present analysis, since only the chord distribution is
considered. The algorithm scheme is reported in Fig. 4.
The adopted optimization tool is the multi-objective genetic
algorithm developed by Mathworks, based on the work of Deb [17]
and implemented as gamultiobj function in the software Matlab
[18]. The algorithm includes the following functionalities:

elitarism selection based on the genotype or phenotype
distance;

roulette wheel parent selection;

random or uniform crossover;

mutation function based on the Gaussian.
A total of 50 individuals are evaluated and evolved for 100
generations. Among the individuals of the rst generation, the
baseline Sandia prole, is included.
The number of decisional parameters np provided by the genetic
algorithm is equal to half the number of vertical divisions. Two
different individual creation methods are tested and implemented
in the WOMBAT algorithm, as shown in Fig. 5. The Troposkien shape
of rotor blade mean line is assumed to be xed, whereas the
decisional parameters are considered as:

Fig. 3. Comparison between experimental data and BE-M predictions (based on the
aerodynamic database provided by Refs. [12,24]) at different angular speeds for the
Sandia 34 m wind turbine [23].

geometry for the given speed. A wind speed equal to 9 m/s is

considered for the proposed computations. The efciency parameter considered in the optimization process is the power coefcient
CP : it can be considered as an efciency of the conversion system
[10] and is dened as

1. independent numerical values of the chord (np genes), dened

as Free Optimization;
2. control point coordinates for a curve (np/2 abscissas and np/2
ordinates) that describe the trend of the chord, dened as
Spline Optimization.
The two methods would ideally provide the same optimal solution but, since the considered number of generations is nite, the
Free Optimization approach does not guarantee a smooth trend,
thus the possibility for a physical realization. On the contrary, the
Spline Optimization approach always considers constructible
geometries. The performance can however be limited because of
the distribution shape constraint.
Traditionally, the main purpose of the optimization activity is to
maximize an efciency parameter for the target operating conditions. The test report by Sheldahl [20] does not deal with the design
of the turbine and, therefore, the target conditions are not clearly
stated. The turbine optimization can be conducted considering any
wind velocity, since the algorithm would provide the optimal

Fig. 4. Functional scheme of the WOMBAT algorithm for wind turbine design (from
[13]).

692

CP

G. Bedon et al. / Energy 66 (2014) 689e698

P
3
0:5$r$A$UN

(1)

being P the power production, r the air density, A the swept area of
the rotor and UN the freestream wind speed.
The wind turbine is assumed to be coupled with a variable speed
generator and would ideally operate at the rotational speed which
maximizes the power coefcient.
The gamultiobj algorithm tends to a solution that minimizes the
tness value. The tness of the prole is thus computed considering
the following expression:

fitness

1
1

CP;Individual;9
max CP;Individual;9;k

(2)

where CP,Individual,9 is the power coefcient of the considered

conguration for a wind speed of 9 m/s and CP,Individual,9,k is the
power coefcient of the considered conguration, computed for a
wind speed of 9 m/s and for the kth rotational speed.
Additional optimization campaigns are successively conduced,
varying the thickness/chord ratio. In fact, the optimization results
from the analysis by Bedon et al. [13] for a blade architecture
characterized by a constant chord distribution, highlighted an
optimal thickness/chord ratio equal to 0.15, differently from
Sandia baseline conguration (which is based on the NACA-0012
airfoil) [20]. Different airfoil proles are therefore additionally
considered, NACA-0015, -0018, -0021 and -0025, assuming a
uniform chord equal to the Sandia turbine case for the baseline
congurations.

4. Optimization results
Fig. 6 shows the optimal chord distribution provided by the
WOMBAT algorithm, considering a thickness-to-chord ratio equal
to 0.12, 0.15, 0.18, 0.21 and 0.25.
The results show that the optimal chord distribution, considering a xed airfoil thickness/chord ratio, is not uniform. The
optimization convergence is not however complete: the resulting
architectures are in fact affected by discontinuities in the
geometrical parameters that will eventually need to be xed
before rotor blade manufacturing. The same trend is approximately registered for all the congurations, presenting a chord
distribution with two (for t/c 0.15 and 0.18) or three (for t/
c 0.12, 0.21 and 0.25) maximum values, at the middle height
(only for t/c 0.12, 0.21 and 0.25) and at two intermediate blade
sections, comprised between the equatorial plane and the blade
tips (for all the considered thickness-to-chord ratios). The curve
stiffness corresponding to the middle height peak can be very
high (especially for t/c 0.12): in the practical realization this
peak needs to be neglected due to static and fatigue concerns
regarding blade geometries characterized by sudden changes in
the section area.
An interpretation for the chord trend is presented as follows:
the minimum value of the chord at the rotor top and bottom leads
to a lower aerodynamic resistance of the blade sections where the
power production is very low; the chord length towards the center
is higher in order to increase the aerodynamic load where the
radius is increasing, thus enhancing the power production; the
successive chord reduction represents a compromise between the
increase of the aerodynamic load and the decrease of the

Fig. 5. Frontal view of the Troposkien rotor blade: the decision variables can be interpreted as independent chord values (a) or coordinates for a spline curve describing the chord
trend over the blade length (b).

G. Bedon et al. / Energy 66 (2014) 689e698

Fig. 6. Baseline (solid line), free optimized (dash-dot line) and spline optimized (light dashed line) chord distributions with respect to total blade height.

693

694

G. Bedon et al. / Energy 66 (2014) 689e698

Fig. 7. Evolution of the power coefcient as a function of the tip speed ratio for the baseline rotor conguration (solid line), the free optimized blade architecture (dash-dot line) and
the spline optimized one (light dashed line).

G. Bedon et al. / Energy 66 (2014) 689e698

interference factor, that would determine an excessive slowdown

of the current if the chord kept on increasing.
As can be clearly seen, slightly different congurations are
registered for the Free Optimization and the Spline Optimization, due to the constraint imposed in the second method. The
constraint to create a spline curve, imposed on the decisional parameters, provides in fact a smoother distribution, thus simplifying
rotor blade construction.
The main performance parameter, considered in order to evaluate the rotor blade performance, is the power coefcient. This is
analyzed with respect to the tip speed ratio, dened as:

uR
UN

(3)

being R the maximum rotor radius.

The comparison between the optimal solutions and the baseline
conguration is shown in Fig. 7.
The power coefcient is increased in all the optimized architectures with respect to the baseline conguration. The Free
Optimization reaches higher peak values than the Spline Optimization, because the chord distribution is not limited by any
shape constraint. However, as explained before, this could prevent
the physical realization of the rotor. Moreover, it can be observed
that the peak is moved towards lower values of the tip speed ratio,
determining a reduction of the optimal rotor angular velocity and,
thus, a decrease of the inertial loads on rotor blades.
Table 2 summarizes the computed values of the peak power
coefcient for the considered blade congurations, as well as the
corresponding angular velocities.
5. Discussion of the numerical results
The performance increase obtained by all the optimized rotor
blade congurations needs to be claried by additional analyses.
For this purpose, the local power coefcient along the blade span
CP,y is computed for every vertical blade subdivision. Local values of
the power coefcient with respect to the blade height are reported
in Fig. 8.
It can be observed that the local power coefcient of the optimized rotors is higher than the baseline conguration for almost
every vertical blade subdivision. In particular, the higher increase in
performance is registered at the center of the blade, where the
radius is maximum. This subsequently leads to a consistent increase in the overall power production in the central portion of the
rotor blade, affecting the overall efciency of the rotor, due to the

695

high lever-arm of such blade stations with respect to the turbine

shaft.
Fig. 9 shows the evolution of instantaneous torque coefcient
with respect to a single rotor blade, dened as:

Ctb

Tb
2 AR
0:5rUN

(4)

where Tb is the torque generated by a single rotor blade, as a

function of its azimuthal position.
It can be observed that the blade torque is sensibly increased,
especially during the upwind portion of the revolution. A peak in
the torque coefcient is registered in the very same angular position where it is generated by the baseline conguration, but the
value is considerably increased (up to three times) for the optimized blade architectures. On the contrary, in the downwind section, the computed torque is almost the same as in the baseline
conguration, leading to similar aerodynamic performance.
The upwind increase in rotor efciency for the optimized blade
architectures can be further highlighted considering the variation
in the interference factor a, dened as:

a 1

Vblade;i
UN

(5)

where Vblade,i is the ow velocity at blade section (i can be up or

down).
The distribution of the interference factors along the swept area
of the rotor is represented in Fig. 10 for the t/c 0.18 baseline and
optimized cases.
The distribution of the interference factor on the swept area of
the rotor is very similar for all the three examined congurations,
but its magnitude is sensibly higher for the optimized architectures,
due to the increased power production: this means that a greater
amount of energy is extracted from the air, decreasing its velocity.
As previously highlighted, the increase in power coefcient is
experienced especially at the equatorial plane of the rotor, where
the radius is maximum. As a matter of fact, a dramatic increase of
the interference factor is registered at the equatorial plane for the
two optimized blade architectures.
As observed in the previous section, being not limited by any
shape constraint, the free optimized blade geometry registers
higher values of the interference factor with respect to the spline
one, due to its higher efciency.

6. Conclusion and future work

Table 2
Peak values of the power coefcient and corresponding rotational speeds for the
baseline rotor conguration and the two optimized blade architectures.
Airfoil prole

Conguration

CP [e]

n [rpm]

NACA 0012

Baseline
Free optimization
Spline optimization
Baseline
Free optimization
Spline optimization
Baseline
Free optimization
Spline optimization
Baseline
Free optimization
Spline optimization
Baseline
Free optimization
Spline optimization

0.3129
0.3484
0.3269
0.3265
0.3701
0.3556
0.3108
0.3481
0.3414
0.2912
0.3445
0.3319
0.2619
0.3727
0.3623

450
400
350
400
350
300
400
300
300
400
300
300
400
300
300

NACA 0015

NACA 0018

NACA 0021

NACA 0025

A variable chord distribution for the blade of a Troposkien vertical axis wind turbine is proposed, leading to increased rotor
performance with respect to the baseline conguration, characterized by a uniform chord distribution along the whole span.
Five optimization campaigns are conducted, considering
different thickness-to-chord ratios for the blade section, determining a similar evolution of the chord length along the blade span.
The registered increment in rotor performance is appreciable and
can reach up to 6% with respect to the considered baseline
conguration. The highest increases are achieved with both NACA
0015 and NACA 0025 proles. Moreover, the performance increase
occurs with a lower optimal angular velocity than the one registered on the original conguration, leading to a reduced blade
loading due to inertial forces. The registered enhancement in rotor
performance is mainly located in the upwind zone, where the
aerodynamic torque is increased up to three times with respect to
the baseline solution. Finally, the performance is strongly increased

696

G. Bedon et al. / Energy 66 (2014) 689e698

Fig. 8. Local values of the power coefcient with respect to the blade height for the baseline rotor conguration (solid line), the free optimized blade architecture (dash-dot line)
and the spline optimized one (light dashed line).

G. Bedon et al. / Energy 66 (2014) 689e698

Fig. 9. Evolution of the torque coefcient relative to a single rotor blade with respect
to its azimuthal position (wind is coming from the upper side of the gure) for the
baseline rotor conguration (solid line), the free optimized blade architecture (dashdot line) and the spline optimized one (light dashed line).

697

Fig. 10. Distribution of the interference factor on the swept area for the baseline, free
optimization and spline optimization congurations, t/c 0.18.

698

G. Bedon et al. / Energy 66 (2014) 689e698

close to the equatorial plane of the rotor, where the higher leverarm allows to maximize the energy extraction.
A reduction in the chord length at the blade tips is registered for
the optimized geometries, determining a decrease in the aerodynamic drag in those blade stations which generate a reduced
contribution to the overall rotor torque, being placed close to the
rotational axis. Approaching to the equatorial plane of the rotor, an
increase in the chord length allows enhancing the torque contribution of the blade sections characterized by higher radius, while
the registered chord reduction at the equatorial plane should be
further investigated, in order to fully understand its contribution to
the energy extraction. It could be assumed that a too high chord
value at the middle height of the rotor determines a relevant
slowing of the air stream, thus dramatically reducing the ux of
kinetic energy through the central sections of the swept area, being
proportional to the cube of the wind velocity. A full CFD campaign
of analysis could enlighten this assumption through the comparison between a rotor blade characterized by a constantly increasing
chord towards the middle of the rotor and a reduced chord
conguration, as proposed in the present work.
Nomenclature
a []
interference factor
A [m2] rotor swept area
c [m]
airfoil chord
CD [] airfoil drag coefcient
CL [] airfoil lift coefcient
CP [] rotor power coefcient
CP,y [] local power coefcient relative to a vertical blade
subelement
Ctb [] torque coefcient relative to a single rotor blade
CP,Individual,9 [] power coefcient of the considered conguration
for a wind speed of 9 m/s
CP,Individual,9,k [] power coefcient of the considered conguration
computed for a wind speed of 9 m/s and for kth
rotational speed.
h [m]
height of a blade element
H [m] rotor total height
n [rpm] rotor angular speed
np [] number of genes
N []
number of blades
P [W]
power produced by the turbine
r [m]
radius of a blade element
R [m]
wind turbine maximum radius
t [m]
airfoil thickness
Tb [Nm] torque generated by a single rotor blade
UN [m/s] wind speed
Vblade,i [m/s] ow velocity at blade section (i can be up or down)
l []
Tip speed ratio
r [kg/m3] air density (assumed 1 [kg/m3])

References
[1] Almohammadi KM, Ingham DB, Ma L, Pourkashan M. Computational uid
dynamics (CFD) mesh independency techniques for a straight blade vertical
axis wind turbine. Energy Sept. 2013;58:483e93.
[2] Glauert H. Airplane propellers. In: Aerodynamic theory. Berlin Heidelberg:
Springer; 1935. pp. 169e360.
[3] Templin RJ. Aerodynamic performance theory for the NRC vertical-axis wind
turbine. Tech. rep. NASA STI; 1974. Recon N 76: 16618
[4] Strickland JH. The Darrieus turbine: a performance prediction model using
multiple streamtubes. Tech. rep. Sandia National Laboratories; 1975.
SAND75e0431
[5] Paraschivoiu I. Double-multiple streamtube model for Darrieus in turbines. In:
Wind turbine dynamics 1981. pp. 19e25.
[6] Paraschivoiu I, Delclaux F. Double multiple streamtube model with recent
improvements. J Energy 1983;7(3):250e5.
[7] Zhang J. Numerical modeling of vertical axis wind turbine (VAWT); 2004.
[8] Claessens MC. The design and testing of airfoils for application in small vertical axis wind turbines. M.Sc. Thesis. Faculty of Aerospace Engineering, Delft
University of Technology; 2006.
[9] Raciti Castelli M, Englaro A, Benini E. The Darrieus wind turbine: proposal for a
new performance prediction model based on CFD. Energy June 2011;36:
4919e34.
[10] Paraschivoiu I. Wind turbine design: with emphasis on Darrieus concept.
Polytechnic International Press; 2002.
[11] Worasinchai S, Ingram G, Dominy R. A low-Reynolds-number, high-angle-ofattack investigation of wind turbine aerofoils. Proc Inst Mech Eng Part J Power
Energy July 2011;225:748e63.
[12] Sheldahl RE, Klimas PC. Aerodynamic characteristics of seven symmetrical
airfoil sections through 180-degree angle of attack for use in aerodynamic
analysis of vertical axis wind turbines. Tech. rep. Sandia National Laboratories;
1981. SAND80e2114.
[13] Bedon G, Raciti Castelli M, Benini E. Optimization of a Darrieus vertical-axis
wind turbine using blade element momentum theory and evolutionary algorithm. Renew Energy Nov. 2013;59:184e92.
[14] Raciti Castelli M, Fedrigo A, Benini E. Effect of dynamic stall, nite aspect ratio
and streamtube expansion on VAWT performance prediction using the BE-M
model. Int J Eng Phys Sci 2012;6:237e49.
[15] Bourguet R, Martinat G, Harran G, Braza M. Aerodynamic multi-criteria shape
optimization of VAWT blade prole by viscous approach. Wind Energy 2007:
215e9.
[16] Carrigan TJ, Dennis BH, Han ZX, Wang BP. Aerodynamic shape optimization of
a vertical-axis wind turbine using differential evolution. ISRN Renew Energy
2012:1e40.
[17] Deb K. Multi-objective optimization using evolutionary algorithms. In: Multiobjective optimization. Hoboken, NJ: John Wiley & Sons; 2001. pp. 13e46.
[18] Matlab.http://www.mathworks.com/help/techdoc/ref/pchip.html; 2012.
[19] Bedon G, Raciti Castelli M, Benini E. Numerical validation of a blade elementmomentum algorithm based on hybrid airfoil polars for a 2-m Darrieus wind
turbine. Int J Pure Appl Sci Technol 2012;12(1):1e6.
[20] Sheldahl RE. Comparison of eld and wind tunnel Darrieus wind turbine data.
Tech. rep. Sandia National Laboratories; 1981. SAND80e2469.
[21] Reis GE, Blackwell BF. Practical approximations to a Troposkien by straightline and circular-arc segments. Tech. rep. Sandia National Laboratories;
1975. Report SAND74e0100.
[22] Strickland JH, Webster BT, Nguyen T. A vortex model of the Darrieus turbine:
an analytical and experimental study. J Fluids Eng 1979;101(4).
[23] Ashwill TD. Measured data for the Sandia 34-meter vertical axis wind turbine.
Tech. rep. Sandia National Laboratories; 1992. SAND91e2228
[24] Gregorek GM, Hoffman MJ, Berchak MJ. Steady state and oscillatory aerodynamic characteristics of a Sandia 0018/50 airfoil. Tech. rep. Aeronautical
and Astronautical Research Laboratory, The Ohio State University; 1989
[25] Worstell MH. Aerodynamic performance of the 17-metre-diameter Darrieus
wind turbine. Tech. rep. Sandia National Laboratories; 1979. Report SAND78e
1737.