Renewable Energy 89 (2016) 231e244

Contents lists available at ScienceDirect

Renewable Energy
j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m/ l o c a t e / r e n e n e

Effect of twist angle on the performance of Savonius wind turbine

Jae-Hoon Lee, Young-Tae Lee, Hee-Chang Lim*
School of Mechanical Engineering, Pusan National University, 2, Busandaehak-ro 63beon-gil, Geumjeong-gu, Busan, 46241, Republic of Korea

article info abstract

Article history: This study aimed to understand the performance and shape characteristics of a helical Savonius wind turbine at various helica l
Received 28 January 2015
angles. The power coefficient (Cp) at different tip speed ratios (TSRs) and torque coefficient (CT) at different azimuths for
Received in revised form
helical blade angles of 0 , 45 , 90 , and 135 were observed under the conditions of a constant projection area and aspect ratio.
30 October 2015 The numerical results discussed in this paper were obtained using an incompressible unsteady Reynolds average
Accepted 6 December 2015 NaviereStokes (k- RNG) model. A numerical analysis in the unsteady state was used to examine the flow characteristics in 1
Available online 20 December 2015 steps from 0 to 360 . In addition, an experiment was performed at a large-scale wind tunnel, and the results were compared
with those of the numerical analysis. Wind speed correction was also employed because of the blockage effect between the
Keywords: wind turbine and wind tunnel. Our results showed that the maximum power coefficient (Cp,max) values in both cases had
Numerical study
similar tendencies for the TSR range considered in this study, i.e. from 0.4 to 0.8, except for the twist angle of 45 . The C p,
Savonius wind turbine
Helical blade max occurred at the twist angle of 45 , whereas it decreased by 25.5% at 90 and 135 . Regarding the CT values at various
Maximum power coefficient Q- azimuths, the results showed that the peak-to-peak values in the profiles for 90 and 135 were less than those for 0 and 45 .
criterion

2015 Elsevier Ltd. All rights reserved.

1. Introduction manufactured, repaired, and maintained. Moreover, no tail or yaw device for
the wind direction is necessary, because the rotor blade is installed vertically
Because of the excessive use of fossil fuels, the world is facing serious to the ground. Furthermore, VAWTs can generate power even at relatively
problems related to energy depletion and environmental pollution. To low wind speeds compared to HAWTs, and they are also easy to install [2].
overcome these problems, many alternatives to fossil fuels have been VAWTs can be classified into two groups: the Darrieus and Savonius types. A
proposed. Among these, renewable energy has drawn much attention because Darrieus tur-bine is a device that uses the lift force generated by an airfoil,
of the significant investments in its research and development by governments whereas a Savonius turbine exploits the drag force. The Savonius wind
and the diverse policies established by governments to extend it to the private turbine, which was invented in 1929, has an inherently simple shape
sector. According to a report published by the [1]; the amount of renewable compared to other types of wind turbines. Therefore, the cost involved in its
energy generated is increasing yearly. In 2012, the amount had grown by development can be lower. Furthermore, it produces less noise and maintains
about 19% from the previous year. The ca-pacity of wind energy in particular stable performance at relatively low wind speeds (see Refs. [3,4]).
has increased compared to other forms of renewable energy. The annual
average growth rate of wind power capacity from 2007 to 2012 was reported
to be about 25%. Recently, a few studies have been conducted on the optimiza-tion of a
VAWT based on an evolvement in the field of experimental study and
numerical analysis [10]. and [11] numerically studied the influence of the
Within wind energy, horizontal axis wind turbines (HAWTs) have overlap ratio of a Savonius wind rotor. The results showed that the maximum
attracted most of the attention during recent years. However, vertical axis performance appears at an overlap ratio of 0.15. Regarding the numerical
wind turbines (VAWTs) have an inherent advantage over HAWTs. For study with the steady Reynolds average NaviereStokes (k- RNG) model,
example, in the case of VAWTs, the blade is easily some recent papers simulated the vertical axis wind turbine rotors (see
[5e9,12,38]); examined the influences of the diameter-to-height aspect ratio of
a Savonius wind rotor and an increase in the number of stages on the
performance. They also analysed the performance of a Savonius
* Corresponding author.
E-mail address: [email protected] (H.-C. Lim).

http://dx.doi.org/10.1016/j.renene.2015.12.012
0960-1481/ 2015 Elsevier Ltd. All rights reserved.
232 J.-H. Lee et al. / Renewable Energy 89 (2016) 231e244

There has been some literature regarding the helical Savonius wind
turbines. (see Refs. [32e35]). However, have focused on their specific cases.
For instance [32]; studied the Savonius-Darrieus turbine model combined
with the k- turbulence model, and they validated their numerical model
through comparison with existing results [34]. attempted to obtain
performance data of a helical Savonius turbine (45 ), and interestingly, they
found a marginal increase in the power coefficient [35]. conducted numerical
and experimental studies on a variety of helical Savonius turbines (45 - 720 ),
but the software platform and test model used were not reliable enough to
support the power performance [33]. proposed a guideline for designing an
appropriate helical Savonius geometry by utilising the calculus principles of
definite integrals, which would help to gain a basic understanding of the
turbine design.

If we take into account existing studies, it is evident that many previous

researchers have focused on studying various shapes of Savonius wind
turbine. However, a closer look at the design pa-rameters clearly shows that
there is a lack of clear analysis results that would indicate the effects of the
helical angle on the perfor-mance of a VAWT. These previous studies
encouraged us to develop and optimise the Savonius wind turbine with
different helical an-gles by means of an experiment and numerical
calculation. Therefore, the primary objective of this study was to investigate
the variation in the power coefficient and flow patterns of a wind turbine at
different helical angles based on a constant projection area, which is the area
of the wind rotor actually receiving the wind. This paper is organized in the
Fig. 1. Top view of Savonius wind turbine design.
following manner: Section 2 outlines the basic description of experimental
and numerical methods with various blade models. Section 3 describes the
wind turbine at a 90 twist angle. They reported that the perfor-mance at a low parametric analysis of VAWTs under uniform wind flow. Section 4 explains
aspect ratio (0.88) was better than those at 0.93 and 1.17 [13,14]. performed a the effect of various blade parameters on Savonius VAWT, and Section 5
numerical analysis on a Savonius turbine with either two or three blades. The gives the major conclusions.
results indicated that the two-bladed rotor generated better power coefficients
than the three-bladed design. Furthermore, they attempted to optimize the
blade shape using evolutionary algorithms [15]. numerically stud-ied the
influences of the number of blades, overlap ratio, twist angle, and aspect ratio
2. Design of wind tunnel experiment
on the power coefficient [16]. conducted an experiment using a helical blade
in a wind tunnel. Their results indicated that an increase in the twist angle
2.1. Definitions of wind turbine performance
enhanced the perfor-mance at low speeds. On the other hand, increasing the
twist angles resulted in a reduction in the net positive torque.
Usually, it is not easy to evaluate the performance of wind rotors with
different shapes using a wind tunnel experiment. In addition, it is time
consuming to fabricate an appropriate measurement

Fig. 2. Top and side view of wind rotor shapes with different twist angle (solid line: contact line on the upper endplate, dashed line: lower endplate).
 J.-H. Lee et al. / Renewable Energy 89 (2016) 231e244 233

Fig. 3. Various projection areas and blade shape of different azimuths.

system. In addition, because of the limited conditions implanted in

the boundary condition, a numerical simulation cannot be applied
effortlessly. Therefore, we utilized both approaches in this study, D 2D H (2)
h
D H
whereas most studies have focused on and used only one. In order
to define the wind turbine performance and evaluate the interac-
where D is the diameter of the Savonius turbine.
tive flow characteristics between the fluid flow and rotational
TSR is defined as the ratio of the blade tip linear speed to the
blades, it is generally important to express the performance using
undisturbed flow speed. TSR can be expressed in Eqn (3), where R
well-known non-dimensional parameters. These parameters Cp, CT,
denotes the rotor radius [m], n is the revolutions per minute [rpm],
and TSR were used. In particular, Cp is a coefficient used to present
the wind rotor performance. and V is the free stream wind speed [m/s].
The Reynolds number based on the configuration of the wind
turbine is expressed as uR 2pRn
TSR V 60V (3)
rVH
Re m (1) The power coefficient Cp is the ratio of the power produced by
the wind rotor to the power available at a specific wind speed. The power
where V and H are the velocity at the tunnel freestream and the height of the coefficient can be calculated using Eqn (4), where T repre-sents the torque
3
Savonius turbine, respectively. In this study, V and H are taken as 8 m/s and [N,m], r is the air density [kg,m ], and A is the area covered by the rotor
10 m/s and 2.1 m, respectively. Therefore, the Reynolds numbers used in this 2
[m ].
6 6
study are 1.8 10 and 1.44 10 for the experiment and simulation, respectively.

In addition, in order to define the power and torque coefficients (Cp and CT,
respectively), the dynamic effects of the rotational wind turbine need to be
considered; therefore, the hydraulic diameter Dh of the Savonius turbine is
used to form an appropriate projection area A.

Table
1
Average, maximum, and minimum projection areas at different twist angles.

Twist angle Average area Maximum area Minimum area

f 0 0.296 m2 0.392 m2 0.136 m2

f 45 0.297 m2 0.386 m2 0.183 m2
f 90 0.297 m2 0.366 m2 0.229 m2
f 135 0.297 m2 0.335 m2 0.268 m2
Fig. 4. Experimental setup used for wind tunnel test.
234 J.-H. Lee et al. / Renewable Energy 89 (2016) 231e244

Fig. 5. Experimental setup used for wind tunnel test.

and lower end-plates of the blades. In order to properly join all the
Tu blades and stabilize the flow around them, the use of both upper
Cp 0:5rAV3 (4)
and lower end-plates was the best choice, as previously suggested
(see Refs. [20,21]). Regarding the aspect ratio, a was set to 1.33:1,
The torque coefficient CT can be calculated using Eqn (5). In a
which was considered to be the optimum shape [19]. Overlap ratio
case where the wind rotor lift is used to determine the rotational
force, the torque can be generated using the moment due to the lift
b was set to 0.167, and end-plates were installed. In order to
conduct an experiment and numerical analysis based on different
produced by the rotating plane of the blade. On the other hand, in
the case of a drag-type wind rotor, the torque is generated using the twist angles, we made four different models: - 0 , 45 , 90 , and
moment due to the drag. 135 .
In the case of a Savonius wind rotor, the projection area would
T change along a cycle of rotation when two blades are rotating.
CT 0:5rARV2 (5) Therefore, when the blades are rotating, a performance evaluation
needs to consider a full cycle. Depending on the twist angle, the
projection area appears to have a variety of shapes: a nut, ellipsoid,
almost circle, etc. At a twist angle of 0 , the projection area has the
shape of two partly overlapping circles. As the twist angle in-
2.2. Design of helical Savonius blade shape
creases, however, the projection area turns into an ellipsoid shape,
as indicated in Fig. 3. Therefore, the various projection area shapes
As shown in Figs. 1 and 2, a variety of parameters are used to
at different twist angles were taken into consideration in this study.
design a helical Savonius wind turbine blade. Some representative
Note that even at different twist angles, the wind turbine was
design parameters are as follows: the aspect ratio (a), overlap ratio
designed to have projection areas with identical average sizes. The
(b), twist angle (f), and azimuth angle (q). The aspect ratio a0 is maximum, minimum, and average projection areas are listed in
defined as the ratio of the height (H) to the diameter (D) of the
Table 1. The figures in this table indicate that the difference be-
blade, as shown in Eqn (6). In order to find the influence of the
tween the minimum and maximum projection areas became lower
aspect ratio, this study considered the effective a0 as one of the as the twist angle increased.
primary parameters, which may increase the rotor performance
efficiency, as reported in previous papers (see Refs. [13,19]).
2.3. Wind tunnel experiment
H
aD (6) The experiments were carried out in a large-scale boundary

Equation (7) represents overlap ratio b0, where b is the ratio of the gap
between two adjacent blades (e) to the distance between both blade ends (D)
[18]. explained the effect of the overlap ratio between blades using particle
image velocimetry (PIV). The parameter e is used to define a gap, and the
oncoming wind blows through the gap along a concave surface of the blade,
which lets air move through this gap and reach an opposite blade. When the
overlap ratio b increases to some extent, the torque and power coefficient
reach their maximums and decrease (see Refs. [38,39]). Hence, overlap ratio
b is one of the design parameters used to in-crease the performance of
Savonius rotor blades. Some papers have already been published on this
subject (see Ref. [10,11,13]). There-fore, the effect of the overlap ratio was
not considered in this study.

e
b D (7)
The twist angle is defined as the twist angle between the upper Fig. 6. Variation of velocity correction factor with S/C [21].
 J.-H. Lee et al. / Renewable Energy 89 (2016) 231e244 235

layer wind tunnel at Pusan National University. The dimensions of the wind reconnects with the external domain so that the wind flows are readjusted and
tunnel were 2 m 2.1 m 20 m. A 185 kW three-phase variable speed DC motor repeatedly renewed. Therefore, after creating this unsteady condition, it is
was used to keep the wind speed constant. The maximum wind speed was finally stabilized. The final values depend on the number of iterations (i.e. 50
limited to 23 m/s with the turbu-lence intensity less than 1% (see Ref. [17]). iterations in our study) and then become converged. In addition, the data
One of authors previ-ously described the experimental method in detail (see began to be saved after five rotations of the turbine rotor to ensure flow
Refs. [22,23]). Therefore, only a short introduction and brief description of stabilization. The data began to be saved after five rotations of the turbine
the experiment is given here. Fig. 4 shows the details of the experimental rotor to ensure flow stabilization.
setup employed. A pitot tube was installed 5 m ahead of the wind turbine to
measure the non-disturbed wind speed in the upstream region. The incoming
wind speed was measured using a micro-manometer (FCO12). 3.1. Governing equation

The turbulence model employed in this paper requires an un-steady

The incoming wind from the wind tunnel rotated the blade and generated Reynolds average NaviereStokes (URANS) analysis. In this case, the
rotating power. The rotational power produced as a result of the wind flow governing equations under a Newtonian fluid condition required two
was analysed using a torque meter (TRD-50KC) built in a circular box placed equations: the continuity equation expressed in Eqn (9) and momentum
below the turbine model. When measuring the torque, we used a powder equation expressed in Eqn (10).
brake (ZKG-50YN) to create force control power. The rotational speed (rpm)
of the wind turbine was measured using a tachometer, which was installed to v ui
receive the signal sent by an optical sensor (ROS-5P). The torque meter was vxi 0 (9)
connected to a computer through an A/D converter, through which the voltage i
vt vxj uiuj r vxi vxj n vxji ui0uj0 ! (10)
signal produced by the torque meter was transmitted to the computer. The vu v 1v p v vu
signal travelling from the torque meter to the computer was acquired by
means of in-house code, which was coded using the Labview platform data
acquisi-tion software. The output power of the helical Savonius wind tur-bine
was calculated based on the measured values. Fig. 5 shows the helical where ui and u0i are the mean and fluctuating components, respectively, of
Savonius wind turbine model and measurement devices used in the velocity in the xi direction. In addition, p is the mean pressure, n is the
experiment. kinematic viscosity, r is the density of the fluid, and t is the time. The
Reynolds stress u0iu 0j also needs to be modelled to close the problem
mathematically. (See Ref. [26] Among the various turbulence models (e.g.,
standard k-u and k-, etc), the k- RNG model was chosen to better predict
2.4. Wall interference effect (blockage effect) the swirling effect behind the rotating blade, particularly to enhance the
accuracy of the rapid strain and streamline curvature (see Refs. [24,26,27].
The Cp of the wind turbine was affected by the wall (see Refs. [21,22]). The turbu-lence kinetic energy (k) is also described in Eqn (11). The
turbulence dissipation rate () is given by Eqn (12).
Many researchers have tried to study this effect. Among them [21]; suggested
a relation for the velocity correction due to the blockage ratio of a Savonius r am !
wind turbine, as shown in Eqn (8) and Fig. 6. Dt vxj k eff vxj Gk r (11)
Dk v vk

Vc 1 r am ! C G C r
S (8) Dt vxj eff v xj 1 k k 2 k R (12)
2
V 1 mC Dk v v
where the blockage ratio (S/C) is the ratio of the wind turbine projection area
(S) and wind tunnel cross-sectional area (C). V is the free stream velocity, Vc
is the correction velocity, and m represents the coefficient of wall, which had where ak and a are the turbulent Prandtl numbers for k and . In addition,

a value of one. In our study, the values of S/C and Vc/V were 0.092 and 1.15, meff and Gk are the dispersion coefficient and the genera-tion of turbulence
respectively. After correction, the velocity in the wind tunnel during our kinetic energy due to the mean velocity gradi-ents, respectively. In these
experiments was increased from 10 m/s to 11.5 m/s, while the velocity in the equations, C1 and C2 are constants having values of 1.42 and 1.68,
numerical simulations was changed from 8 m/s to 9.2 m/s.
respectively. In addition, the term R is used to improve the accuracy for
rapidly strained flows (see Refs. [24,26,27].

3. Numerical analysis A k- RNG turbulence model was selected for our analysis. In

In this study, the numerical simulations were coaxially per-formed using

ANSYS Fluent, which is a commercial computational fluid dynamics (CFD)
solver. This software calculates the compli-cated flow structure based on the
finite volume method (FVM) of the NaviereStokes governing equation, which
is suitable for resolving the problems associated with the interaction between
the complicated on-coming wind flow and the rotating blades. The numerical
domain and meshes were generated using ANSYS ICEM. The number of
meshes used in this study ranged from 1,200,000 to 1,500,000.

In order to calculate the flow around the wind turbine, it is important to set
an appropriate iteration time at each step during the rotation of the blade. As
the subdomain is rotated in each step, it Fig. 7. Overall domains of boundary and internal condition used in numerical analysis.
236 J.-H. Lee et al. / Renewable Energy 89 (2016) 231e244

order to deduce the link between the pressure and velocity in the calculation 3.2. Boundary conditions
domain, the semi-implicit method for pressure-linked equation (SIMPLE)
algorithm was used. For appropriate analysis, the overall domain was divided into two sub-
In order to deduce the link between the pressure and velocity in the domains: surrounding fixed and inner rotating bladed domains as shown in
calculation domain, we implemented a second-order upwind scheme (thus Fig. 7. The total number of grids was 1.0~ 1.5 million, and the grid shape is
improving accuracy and feasibility) rather than using a first-order scheme. In shown in Fig. 8. Fig. 8 (a) and (b) are the main grid shapes of the rotating
order to convert the continuity equation into a discrete Poisson equation for rotor and surrounding outer domain, respectively. In order to link the inner
pressure, the Simple method was applied (see Refs. [11,14,27]). The and outer domains, the interface condition was used to describe the separated
differential equations are linearized and solved implicitly in sequence, starting wake flow interaction with the rotating blades and surrounding region. In
with the pressure equation (predictor stage), followed by the momentum addition, the sliding mesh model (SMM) was used for a (pseudo-) rotating
equations and the pressure-correction equation (corrector stage). In order to mesh to simulate the rotating blades. The sliding mesh could be effectively
manipulate the gradient, we used a least-squares cell-based scheme. (See Ref. used in a case where the mesh did not deform. The rotational speed could be
[28]). set depending on the experimental conditions (see Refs. [11,24,27]).

4
The time step we used in the calculation was 5.89 10 sec at TSR1 (i.e.
1 rotation every time step) to observe the detailed structure of the separated In order to impose a similar condition as the wind tunnel, the inlet
wake behind the turbine blades. For a reliable result, the calculation was boundary conditions were set as follows: velocity inlet at a uniform velocity
continuously made to achieve consistent torque from each blade during one of 8 m/s, and the outlet atmospheric pressure condition at 1 atm. The no-slip
cycle. In addition, in order to provide a suitable time step, the CFL number wall condition was applied to the surface of the domain wall and the blade
was main-tained at less than 10, which is a bit unsuitable, but the standard surface. In terms of tur-
wall function compensates for the wall treatment instead. 1 0 0 1 0 2 0 2 0 2
bulent kinetic energy (k), k is defined as 2 u iu j 2 u x u y u z
3 02
2 u and the axial stresses are assumed to be approximately 0.7 in our
During each iteration, the values obtained for the variables should get
study. Therefore, the turbulent kinetic energy (k) and turbulent dissipation
closer and closer so that they converge. For some reason, the solution can rate () are defined as a unit. The wall boundary con-dition was applied to the
become unstable, so a relaxation factor refers the value from the previous side and top/bottom wall planes. The moving wall condition was set for all
iteration to dampen the solution and cut out steep oscillations. As a rule of the moving components such as the helical blades, main supporting pipe, and
thumb in this study, we simply keep the relaxation factors at default, which is end plates. In order to observe the vortex formation behind the blades in
quite reasonable for especially cold flows without combustion. In our study, detail, the downstream size of the sub-domain was set at around 4D.
we used pressure 0.3, body force 1, momentum 0.8, turbulent kinetic energy
0.8, turbulent dissipation rate 0.8, and turbulent viscosity 1. Regarding the sub-domain side, this study mainly attempts to understand
the near-vortex flow close to the blades determining the fine subdomain in the
downstream approximately 4D, which

Fig. 8. Mesh generation and distribution around VAWT.

 J.-H. Lee et al. / Renewable Energy 89 (2016) 231e244 237

simulation results at different twist angles (f) ranging from 0 to 135 .

Because of the blockage effect, Cp,max was reduced to 38.2% in the

experiments and numerical simulations. Compared to 0 and 45 , Cp appeared
relatively low at twist angles of 90 and 135 . C p,max was found to be
between TSR values of 0.5 and 0.65 for all of the twist angles. In addition, the
Cp values were not present below a TSR of 0.4 in the experiment, which was
due to the high mechanical friction between the main axle of the Savonius
turbine and the mechanical powder brake. The difference between the Cp,max
values in the experimental and simulation results was the largest (around
0.02) at the 45 twist angle, whereas at the 135 twist angle, this difference was
found to be very small (i.e. almost negligible). The performance of the
Savonius wind turbine was observed to be the most efficient at the 45 twist
angle from both the experiments and numerical analysis, with a Cp of 0.13 at
a TSR of 0.54, whereas the twist angle with the lowest value of Cp was 135 ,
with a Cp of 0.12 at a TSR of 0.54.
All experimental data may contain more or less uncertainty. An
uncertainty analysis was carried out for all experimental results to assess their
confidence levels, following the method suggested by Ref. [25]. The total
error consists of the bias error and precision error. The bias error can be
minimized by carefully calibrating the measuring instruments. To evaluate the
precision error, the stan-dard deviation of the sample records was calculated
for the surface pressure. The total error, with 95% confidence, is depicted as a
form of error bar. (see Fig. 9) As shown in the figure, the maximum error of
uncertainty reaches approximately 5% at most.

4.2. Temporal variation of torque coefficient at different azimuths

Fig. 10 shows the torque coefficient (CT) values at different azi-muths.

When the azimuth was varied, CT attained its highest value of 0.34 at a twist
angle of 45 and TSR of 0.45. The graphs also indicate that the phase
Fig. 9. Power coefficient variations against TSR.
difference of CT decreased as the twist angle increased. At a twist angle of
135 , the phase difference was the least.
includes the smallest scales of the vortex. In fact, a parametric study was also
conducted for the downstream subgrid of the wind blades. It was found that In the experiment, the torque sensor actually reads the averaged torque
an approximately 4D to 6D subdomain in the downstream of the blades was values during the measurement so that the values of different azimuth angles
enough to generate small-scale eddies. In addition, in order to maintain the in real time may not be possible or available owing to the hardware
vortex in the region that is farther downstream, the midsized subdomain was limitations of our experiment. In addition, depending on the condition of the
also created so that the vortex region was still maintained farther downstream. wind tunnel, the torque signal sometimes becomes unstable in the early stages
One of the interesting facts in the result is that the dissipation of the vortex of mea-surement. Therefore, during each measurement, we waited to obtain a
wake behind the blades depends on the twist angle (e.g., see Fig. 12, and the stable condition that yielded a reliable rpm and torque. The averaged values
periodic regular vortex wake appears downstream, whereas the dissipation were obtained after waiting for approximately 5e10 min to get reliable values
seems faster when the twist angle increases). in the tunnel. For the numerical simulation, the torque variation having
consistent periodic values from each blade was averaged for each cycle.

In the result, negative CT values occurred in the azimuth angle ranges of

4. Results and discussion
60 -150 and 240 -330 at twist angles of 0 , 45 , and 135 , with a TSR of 0.88.
To analyze the results, we compared the Cp values with various TSRs. In Regarding this observation, which will be explained shortly, it is inferred that
a force by air is not properly transferred to the concave surfaces of both
this study, the wind tunnel experiment results were compared with those of
blades. Instead, it might affect convex surfaces. However, in a case where
the numerical simulations. Based on the numerical analysis, C T was TSR was less than 0.45, no negative values were found. This might be
examined at various azimuths under the condition that the blade was being because the air resistance increased at the convex part of the blade in
rotated. Furthermore, we also investigated how the air flow changed at comparison to the rotational power of the turbine as the rotational speed
different azimuths for the Savonius wind turbine. increased. In contrast, the blades with twist angles of 135 did not show any
negative value of CT at TSR 0.88.

4.1. Power coefficients of experimental and numerical analysis

Fig. 11 shows CT values with a TSR of 0.6 at different azimuths.
6 6 Interestingly, in Fig. 11(a) and (b), the CT of blade 1 tends to increase and
The Reynolds numbers used in this study were 1.24 10 and 1.55 10
decrease within the range of 225 -270 . However, this phe-nomenon was not
depending on the averaged projection area. Fig. 9 shows the TSR versus Cp observed in the case of blades with twist angles of 90 and 135 . It seems that
plot for both the experimental and numerical the twist angle caused the internal
238 J.-H. Lee et al. / Renewable Energy 89 (2016) 231e244

Fig. 10. Torque coefficient variations against TSR.

flow to circulate effectively. In addition, with twist angles of 0 and 45 , the the opposite side of the blade. When the twist angle was 90 , the instant kink
internal flow through the central overlap hole stuck the main axis pipe during of the torque coefficient did not appear because of the twist angle. In addition,
the rotation of the blade, and then moved to with an increase in the twist angle, the
 J.-H. Lee et al. / Renewable Energy 89 (2016) 231e244 239

Fig. 11. Torque coefficient variations of Blades 1 and 2 at different azimuths.

convex surface side always faced the on-coming wind, which implied that the generated torque would remain consistent as the
reduced the torque coefficient. From this observation, it can also be twist angle increased.
240 J.-H. Lee et al. / Renewable Energy 89 (2016) 231e244

Fig. 12. Q-criterion distribution around Savonius wind turbine (Q-value was 0.011).

4.3. Flow visualization at different azimuths surface of blade 2, and then hit the concave surface of blade 1. Figs. 13(c) and
14(b) show the air-flow pattern at an azimuth angle of 90 . Looking at the
Fig. 12 shows the vortex formation at a twist angle of 0 and azimuth angle streamlines shown in Fig. 13 and velocity vector field in Fig. 14, it can be
of 0 , 45 , 90 , and 135 . We used the Q-criterion method, which can be seen that the air moves towards the concave surface, and the flow separates at
defined as the flow regions with a positive second invariant of the velocity the inner and outer re-gions. Subsequently, the air moves towards the opposite
(see Refs. [29e31]). The value for the Q-criterion coefficient was set as 0.011 blade, passing through a narrow space between the shaft and the blades. At
in this case. In the case of the Savonius wind turbine, vortex formation was in this moment, the air entering blade 1 hits the concave surface of blade 2,
the direction of rotation. As shown in the figure, the symmetric vortex pairs which creates the rotational power for blade 2. On the other hand, Figs. 13(b)
are separated from both sides of the blade end and propagated downstream, and (d) suggest that the air moves towards the concave part of the blade. It is
yielding horseshoe-shaped vortex structures. In this visualization scheme, it is also found that when the concave surface of the blade receives the force of the
noticeable that the vortex is a bit complicated, but owing to the end plates air, this force is transferred to the concave surface of the opposite blade.
they would be making better stable wake shape.

In Fig. 14(a), the air surrounding the shaft slowly moves from the concave
Fig. 13 shows the case with a twist angle of 0 and a TSR of 0.6 at surface of blade 2 to the concave surface of blade 1. In contrast, Fig. 14(b)
different azimuths. At an azimuth angle of 0 , the results show that the air did shows that the air moving from blade 2 to the inside of blade 1 moves
not directly impact on the concave surface of the blade, as shown in Figs. relatively fast. Moreover, the area marked by the black circle in Fig. 14 is the
13(a) and 14(a). Instead, it impacted on the convex point where the separated eddies

Fig. 13. Streamline and speed magnitude contours at different Azimuths.

 J.-H. Lee et al. / Renewable Energy 89 (2016) 231e244 241

Fig. 14. Snapshots of velocity vector fields at different azimuths.

occur in the direction indicated by the arrow. suction side has little effect on the pressure-driven flow.
Fig. 15 shows the blade surface pressure (Cpr) and velocity vector fields at Fig. 17 shows the surface pressure distribution on the blade surfaces with
twist angles of 0 and 135 at an azimuth angle of 45 . Fig. 15(a) shows the different twist angles. For a twist angle of 0 (see Fig. 17(a)), the surface
vector field around the blade. In the figure, the wind directions are toward the pressure close to the end plates experi-ences little change, but as the twist
page. It can be seen that the surface of the blade is divided into two parts, i.e. angle increases, the surface pressure increases gradually at the bottom area
right and left. Moreover, the surface pressure is the same at each part. In the close to the end plate and reaches a maximum at a twist angle of 135 (see Fig.
case of Fig. 15(b), we can see that unlike those at a twist angle of 0 , here the 17(d)). In addition, in Figs. (a) and (b), it is noted that the surface pressure
directions of the velocity vectors are not only towards the left and right but coefficient at the concave surface has an almost constant distribution, whereas
also in the upward and downward directions. In addition, the pressure fields it increases substantially at the convex surface.
are different in the upper and lower portions of the blade. Generally, the
pressure has been found to increase near the end plate. This seems to occur as
a result of the wall interference effect by the end plate. Fig. 18 shows the sectional averaged pressure distribution around blades
with different twist angles. For slice S1, the overall pressure distribution is a
bit higher than that of the other slices. The implication of this figure is that in
Fig. 16 presents and compares the surface pressure distributions for the the case of twist angle 0 , the oncoming wind impacts the blade at a
twist angles having the highest and lowest performances. For the azimuth perpendicular angle directly so that the surface pressure on the blade is almost
angle having the highest performance, Fig. 16 shows that the pressure side of consistent along the lateral direction through the blade. However, as the twist
the blade (i.e. concave surface) creates a pressure-driven flow on the suction angle changes, the wind is in multiple directions (i.e. horizontal and vertical)
side of the blade (i.e. convex surface). By contrast, for the case having the along the blade surface, and the suction pressure increases (i.e. the colour
lowest per-formance, the pressure side maintains high pressure, whereas the turns brighter at high twist angles). In addition, the

Fig. 15. Surface pressure (Cpr) and velocity field at twist angles of 45 and 135 with an azimuth angle of 45 . In the figure, the wind directions are toward the page.
242 J.-H. Lee et al. / Renewable Energy 89 (2016) 231e244

Fig. 16. Sectional pressure distribution indicating the maximum and minimum Cpr. In the figure, (a) and (c) are the case of maximum Cpr and (b) and (d) the minimum Cpr.

Fig. 17. Averaged surface pressure distribution for different twist angles.
 J.-H. Lee et al. / Renewable Energy 89 (2016) 231e244 243

Fig. 18. Sectional averaged pressure distribution around the blades with different twist angles.

stagnant pressure close to the end plates is a bit high for the blade with a twist 4) The maximum CT was observed at an azimuth angle of 45 and twist angle
angle of 0 (see the slices S1 and S5. The suction pressure would be high close of 0 but varied with the azimuth and twist angles.
to the end plate). However, this effect seems to be reduced for the blade with 5) Regarding the surface pressure distribution around the blade, when the
a twist angle of 135 (see the slice S5). Considering the structural stability, this convex blade faced the flow, the surface pressure had the maximum
effect reduces relatively the vertical load (vertical lift force and bending distribution, while the concave blade had the minimum. While the blades
moment, etc.) on the main rotational axis (i.e. the negative lift force increases were rotating, some sections had an effective torque, and others had a
as the twist angle increases). relative drag force, which retarded the blades rotation.

Acknowledgements
5. Concluding remarks
This work was supported by the Human Resources Develop-ment of the
This study investigated the performance and shape character-istics of a Korea Institute of Energy Technology Evaluation and Planning (KETEP)
helical Savonius wind turbine at various twist angles. The power coefficient grant funded by the Korea government Ministry of Knowledge Economy (No.
(Cp) values at different TSRs and torque coeffi-cient (CT) values at different 20124010203230, 20114010203080). In addition, this research was supported
azimuths for twist blade angles of 0 , 45 , 90 , and 135 were observed under by Basic Science Research Program through the National Research
the condition that the projection area and aspect ratio were constant. The key Foundation of Korea(NRF) funded by the Ministry of Education, Science and
conclusions are summarized as follows. Technology(2013005347).

This research was also supported by the Fire Fighting Safety & 119
1) The simulation results successfully verified the experiment re-sults at a Rescue Technology Research and Development Program funded by the
range of TSRs and maximum power coefficient (Cp,max) values as the Ministry of Public Safety and Security (MPSS-2015-80).
Savonius wind turbine blade twist angle was varied.
References
2) The maximum Cp appears to be approximately 0.13 at a twist angle of 45 .
However, at twist angles of 90 and 135 , the value of the power coefficient [1] Renewable Energy Policy Network for the 21st century, Renewable 2013 Global Status
(Cp) became lower than that at 0 , but the maximum Cp appeared to be Report, 2013.
[2] M. Islam, D.S.K. Ting, A. Fartaj, Aerodynamic models for Darrieus-type straight-bladed
similar.
vertical axis wind turbines, Renew. Sust. Energ. Rev. 12 (4) (2008) 1087e1109.
3) When the twist angle was greater than 90 , it was found that the torque
coefficients stabilized and remained constant. [3] E. Sandra, Evaluation of different turbine concepts for wind power, Renew.
244 J.-H. Lee et al. / Renewable Energy 89 (2016) 231e244

Sust. Energ. Rev. 12 (5) (2008) 1419e1434. [20] U.K. Saha, S. Thotla, D. Maity, Optimum design configuration of Savonius rotor through
[4] S.H. Yoon, H.C. Lim, D.K. Kim, Study of several design parameter on multi blade wind tunnel experiment, J. Wind. Eng. Ind. Aero. 96 (2008) 1359e1375.
vertical axis wind turbine, Int. J. Precis. Eng. Man. 14 (5) (2013) 831e837.
[21] A.J. Alexander, Wind tunnel tests on a Savonius rotor, J. Wind. Eng. Ind. Aero. 3
[5] Y. Chen, Y. Lim, Numerical investigation of vortex dynamics in an H-rotor vertical axis (4) (1978) 343e351.
wind turbine, Eng. Appl. Comput. Fluid Mech. (2015) 1e12. [22] I. Ross, A. Altman, Wind tunnel blockage corrections: review and application to Savonius
[6] H. Beri, Y. Yao, Double multiple stream tube model and numerical analysis of vertical vertical-axis wind turbines, J. Wind. Eng. Ind. Aero. 99 (2011) 523e538.
axis wind turbine, Energy Power Eng. 3 (2011) 262e270.
[7] Y.T. Lee, H.C. Lim, Numerical study of the aerodynamics performance of a 500W [23] M. Takao, H. Kuma, T. Maeda, Y. Kamada, M. Oki, A. Minoda, As straight bladed
Darrieus-type vertical axis wind turbine, Renew. energy 83 (2015) 407e415. vertical axis wind turbine with a directed guide vane row effect of guide vane geometry
on the performance, J. Therm. Sci. 18 (1) (2009) 54e57.
[8] S. Roy, U. Saha, Review on the numerical investigations into the design and development [24] J. Yao, J. Wang, W. Yuan, H. Wang, L. Cao, Analysis on the influence of tur-bulence
of Savonius wind rotors, Renew. Sustain. Energy Rev. 24 (2013) 73e83. model changes to aerodynamic performance of vertical axis wind turbine, Procedia Eng.
31 (2012) 274e281.
[9] N.K. Sarama, A. Biswas, R.D. Misra, Experimental and computational evalua-tion of [25] H.W. Coleman, Experimentation and Uncertainty Analysis for Engineers, Wiley, New
Savonius hydrokinetic turbine for low velocity condition with com-parison to Savonius York, 1989.
wind turbine at the same input power, Energy Convers. Manag. 83 (2014) 88e98. [26] A. Escue, J. Cui, Comparison of turbulence models in simulating swirling pipe flows,
Appl. Math. Model. 34 (10) (2010) 2840e2849.
[10] N. Fujisawa, On the torque mechanism of Savonius rotors, J. Wind. Eng. Ind. Aero. 40 [27] R. Howell, N. Qin, J. Edwards, N. Durrani, Wind tunnel and numerical study of a small
(1992) 277e292. vertical axis wind turbine, Renew. Energy 35 (2010) 412e422.
[11] J.V. Akwa, G. Alves da Silva Junior, A.P. Petry, Discussion on the verification of the [28] A. Gerasimov, Modeling Turbulent Flows with FLUENT, ANSYS, Europe, 2006.
overlap ratio influence on performance coefficients of a Savonius wind rotor using [29] B.E. Launder, D.B. Spalding, The numerical computation of turbulent flow, Comput.
computational fluid dynamics, Renew. Energy 38 (1) (2012) 141e149. Method Appl. M. 3 (2) (1974) 269e289.
[30] P. Chakraborty, S. Balachadar, R.J. Adrian, Kinematics of local vortex identifi-cation
[12] M.A. Kamoji, S.B. Kedare, S.V. Prabhu, Experimental investigations on single stage, two criteria, J. Vis. 10 (2) (2007) 137e140.
stage and three stage conventional Savonius rotor, Int. J. Energ. Res. 32 (2008) 887e895. [31] C.R. Hunt, A.A. Wary, P. Moin, Eddies, streams, and convergence zones in turbulent
flows, Stud. Turbul. Using Numer. Simul. Databases (1988) 193e208.
[13] M.A. Kamoji, S.B. Kedare, S.V. Prabhu, Performance tests on helical Savonius rotors,
Renew. Energy 34 (2009) 521e529. [32] B.K. Debnath, A. Biswas, R. Gupta, Computational fluid dynamics of a com-bined three-
[14] M.H. Mohamed, G. Janiga, G. Pap, D. Thevenin, Optimization of Savonius turbines using bucket Savonius and three-bladed Darrieus rotor at various overlap conditions, J Renew.
an obstacle shielding the returning blade, Renewble Energy 35 (2010) 2618e2626. Sustain. Energy 1 (2009) 033110.
[33] N. Halsey, Modeling the Twisted Savonius Wind Turbine Geometrically and Simplifying
[15] Z. Zhao, T. Zheng, X. Xu, W. Liu, G. Hu, Research on the Improvement of the its Construction, Oregon Episcopal School, 2011.
Performance of Savonius Rotor Based on Numerical Study, in: SUPERGEN 09. Int. [34] B. Deb, R. Gupta, R.D. Misra, Performance Analysis of a Helical Savonius Rotor without
Conference, 2009, pp. 1e6. Shaft at 45 Twist Angle Using CFD, vol. 7, 2013, pp. 126e133.
[16] U.K. Saha, M.J. Rajkumar, On the performance analysis of Savonius rotor with twisted [35] L. Duffett, J. Perry, B. Stockwood, J. Wiseman, Design and evaluation of twisted
blades, Renew. Energy 31 (2006) 1776e1788. Savonius wind turbine, Vert. wind energy Eng. (2009) 1e33.
[17] K.C. Kim, S.K. Kim, S.Y. Yoon, PIV measurements of the flow and turbulent [38] S. Roy, U. Saha, Review on the numerical investigations into the design and development
characteristics of a round jet in crossflow, J. Vis. 3 (2000) 157e164. of Savonius wind rotors, Renew. Sustain. Energy Rev. 24 (2013) 73e83.
[18] N. Fujisawa, F. Gotoh, Visualization study of the flow in and around a Savonius rotor,
Exp. Fluid 12 (1992) 407e412. [39] O.B. Yaakob, K.B. Tawi, D.T.S. Sunato, Computer simulation studies on the effect over
[19] H.B. Yang, Wind Tunnel Study on the Performance Characteristics of Savonius Wind lap ratio for Savonius type vertical axis marine current turbine, Int. J. Eng. Trans. A
Turbine [Master thesis], Pusan National University, 2013. Basics 23 (2010) 79e88.