Renewable Energy 35 (2010) 2618e2626

Contents lists available at ScienceDirect

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

Technical Note

Optimization of Savonius turbines using an obstacle shielding the returning blade

M.H. Mohamed, G. Janiga, E. Pap, D. Thévenin*
Lab. of Fluid Dynamics and Technical Flows, University of Magdeburg “Otto von Guericke”, Germany

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

Article history: Due to the worldwide energy crisis, research and development activities in the field of renewable energy
Received 16 March 2009 have been considerably increased in many countries. In Germany, wind energy is becoming particularly
Accepted 8 April 2010 important. Although considerable progress has already been achieved, the available technical design is
Available online 8 May 2010
not yet adequate to develop reliable wind energy converters for conditions corresponding to low wind
speeds and urban areas. The Savonius turbine appears to be particularly promising for such conditions,
Keywords:
but suffers from a poor efficiency. The present study considers a considerably improved design in order
Savonius rotor
to increase the output power of a Savonius turbine with either two or three blades. In addition, the
Wind energy
Optimization
improved design leads to a better self-starting capability. To achieve these objectives, the position of an
Evolutionary algorithms obstacle shielding the returning blade of the Savonius turbine and possibly leading to a better flow
orientation toward the advancing blade is optimized. This automatic optimization is carried out by
coupling an in-house optimization library (OPAL) with an industrial flow simulation code (ANSYS-
Fluent). The optimization process takes into account the output power coefficient as target function,
considers the position and the angle of the shield as optimization parameters, and relies on Evolutionary
Algorithms. A considerable improvement of the performance of Savonius turbines can be obtained in this
manner, in particular a relative increase of the power output coefficient by more than 27%. It is
furthermore demonstrated that the optimized configuration involving a two-blade rotor is better than
the three-blade design.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction the earth’s wind. Greenpeace predicted that about 10% electricity
could be supplied by the wind in 2020 using improved technology;
Wind energy is one of the most promising sources of renewable experts predict that wind power would capture 5% of the world
energy. It is pollution-free, abundantly available in the earth’s energy market by the year 2020 [1].
atmosphere, can be locally converted, and can thus help in reducing The storage and distribution of electrical power is still a major
the dependency on fossil fuels. Many developed and developing problem, in particular when the generated quantity is varying
countries have realized the importance of wind as an important considerably with time and location, like the case for wind energy.
resource for power generation and necessary measures are being A local electricity production, within urban areas, would help solve
taken up across the globe to tap this energy for its effective utili- this issue. The Savonius turbine appears in principle to be partic-
zation in power production. ularly promising for such conditions, since it is a slow-running
According to the results released by the Global Wind Energy machine with a very compact design.
Council (GWEC), wind energy developing countries (more than 70) As a main target application, optimized Savonius turbines would
have taken the net wind energy installed capacity to a record high therefore be perfectly suited for electricity production at the level
exceeding 742 GW. Despite constraints facing supply chains for of single buildings with a flat roof, but could also be useful for
wind turbines, the annual market for wind energy continues to instance for off-shore platforms as well as for automatic marine
increase at a staggering rate higher than 30%. It has been estimated stations gathering climate data.
that roughly 10 million MW of energy are continuously available in Savonius initially developed the vertical axis Savonius rotor in
the late 1920s. The concept of the conventional Savonius rotor is
based on cutting a cylinder into two halves along the central plane
and then moving the two half cylinders sideways along the cutting
* Corresponding author. Tel.: þ49 391 67 18570; fax: þ49 391 67 12840.
plane, so that the cross-section resembles the letter S (Fig. 1).
E-mail addresses: [email protected] (M.H. Mohamed), janiga@
ovgu.de (G. Janiga), [email protected] (E. Pap), [email protected] The Savonius rotor, which is a slow-running vertical axis wind
(D. Thévenin). machine (typically used for lx1:0 or below, see later Eq. (1)) has

0960-1481/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.renene.2010.04.007
 M.H. Mohamed et al. / Renewable Energy 35 (2010) 2618e2626 2619

Using the notations of Fig. 2, the velocity coefficient (or speed

ratio) is first defined as:

uR
l ¼ (1)
U
Considering a Savonius rotor of height H, a wind of incoming
velocity U, the mechanical power P and the mechanical torque M on
the axis of a Savonius turbine, it is possible to write:

P
Cp ¼ (2)
rRHU 3
Fig. 1. Conventional Savonius rotor.
and
unfortunately a rather poor efficiency when considering the stan-
dard design: theoretically, Cp x0:18 at best. Nevertheless, it pres-
M
Cm ¼ (3)
ents many advantages for specific applications, in particular due to rR2 HU 2
its simplicity, resulting robustness, compactness and low cost. If where Cp and Cm are respectively the output power coefficient and
a higher efficiency could be obtained, the Savonius rotor would the torque coefficient of the turbine. In the following sections, the
become a very interesting complementary source of electricity optimization process will consider the output power coefficient Cp
from wind energy. as target function, but only for a conventional Savonius geometry.
This means that the geometrical parameters a and e are always
fixed and respectively equal to 0 and R/3. For the three-blade
2. Purpose of the present work configuration, the additional geometrical parameter is given by s/
R ¼ 0.1. This conventional configuration of the rotor has been
With the objective of increasing the efficiency of Savonius extensively studied in the past (see e.g., list of citations in [8,12]).
turbines, a comprehensive review of the corresponding literature The corresponding values of Cp and Cm have been determined
has been first started. The poor efficiency of the standard design is numerically and sometimes experimentally as a function of the
well-known and documented in a number of studies, e.g., [2e5]. speed ratio l. This will be used later to validate the numerical
Therefore, many modifications have already been proposed to procedure by comparison with published, reference data.
improve the performance of this turbine (see in particular refs.
[5e11]). Some of the most important results concerning such
possible improvements are summarized in Table 1. 3. Optimization procedure
Since one of the major advantages of the Savonius turbine is its
simplicity and corresponding compactness, robustness and low Until recently, the denomination “optimization” was mostly
cost, a modification introducing a high complexity should probably used in the engineering literature to describe a trial-and-error,
not be retained. Considering the results of the previous studies
listed in Table 1, some simple guiding or deflecting plate would lead
to the best efficiency improvement at the lowest possible cost and
complexity. Therefore, we will now investigate numerically the
effect of an obstacle shielding partly the returning blade of a Savo-
nius turbine. This work builds on top of a previous investigation
[11] by considering Savonius turbines with two as well as with
three blades and by quantifying the resulting progress in more
detail. Adding a shielding obstacle should in principle reduce the
reverse moment, and as a consequence the total moment of the
turbine will be increased, since the total moment is the moment
difference between the advancing and the returning blades. Related
ideas have already been proposed by other groups in the past (see
again Table 1). But, at the difference of the previous studies, we are
not looking here for a better solution, but directly for the best
possible one. More specifically, the position and the angle of the
shielding obstacle will be optimized, thanks to a corresponding
mathematical procedure. For that purpose, it is first necessary to
define key parameters, suitable to measure the efficiency of the
Savonius turbine.

Table 1
Possible modifications to improve the performance of Savonius turbines.

Design modification Gain Description and comments

Helical rotors [5] Improve the static torque Complex design, high cost
Deflector Plate [6] 20% No further details since 1992
Twisted-blade [7] 27% relative Complex design, high cost
Guide-Box Tunnel [8] 50% (3 blades) Complex design
Modified Savonius [9] 60% in static torque Expected vibration problem
Guide Vanes [10] Depends on wind speed Problems for large l
Fig. 2. Schematic description and main parameters characterizing a Savonius rotor
Obstacle Plate [11] 15% on peak value Small parameter space used
with two or three blades.
2620 M.H. Mohamed et al. / Renewable Energy 35 (2010) 2618e2626

manual procedure at the difference of a real, mathematical opti- the flow equations obtained by Computational Fluid Dynamics
mization. This is now changing rapidly. In the present project, the (CFD) [16]. In the present paper, the free design variables consid-
optimal position and angle of the obstacle will rely on a mathe- ered for the optimization will describe the obstacle position. For
matical optimization, as a first step toward an even more aggressive this purpose, three parameters are considered (X1, Y1 and X2) which,
optimization of all important parameters affecting the performance together with a fixed value for Y2 are sufficient to fix clearly the
of the different Savonius turbines, also for non-conventional geometry of the shielding obstacle (see Fig. 3). The objective
parameters a, e, s. function considers only one output of the simulation, that should
The mathematical optimization of turbomachines is still a rela- be maximized as far as possible: the output power coefficient Cp.
tively new field or research (see for instance refs. [13e15]). Opti- An appropriate algorithm must then be chosen to carry out the
mization is used here to identify the best position and angle of the computations. A considerable experience is available in our group
obstacle shielding the returning blade. The central goal when concerning the mathematical optimization relying on CFD-based
designing an improved Savonius turbine is to achieve high effi- evaluations [16]. We therefore employ our own optimization
ciency, i.e., high power output. Furthermore, it must be kept in library, OPAL (for OPtimization ALgorithms), containing many
mind that turbomachines often operate outside the nominal (or different optimization techniques. OPAL has already been coupled
design) conditions. Therefore, after optimizing the configuration in the past with different CFD solvers (in-house codes, ANSYS-
for the maximum output power coefficient, known to occur for Fluent, ANSYS-CFX) and has already been employed successfully to
a speed ratio lx0:7, the full range of speed ratios will be improve a variety of applications, for example heat exchangers [17],
considered. turbomachines [11,18] or burners [19,20]. These studies have in
Mathematical optimization methods attempt to determine the n particular demonstrated the efficiency of Evolutionary Algorithms
design variables Xi(i ¼ 1.n) that maximize a user-defined objective (EA) for CFD-based optimization. In this manner, a very robust
function, denoted OF(V(Xi), Xi), where V(Xi) is here the solution of solution can be obtained and local extremal values do not falsify the

Fig. 3. Schematic description of the geometry and free optimization parameters X1, Y1, X2 used to modify the position of the shielding obstacle. Top: two-blade Savonius rotor;
bottom: three-blade Savonius rotor.
 M.H. Mohamed et al. / Renewable Energy 35 (2010) 2618e2626 2621

Table 2
Parameters of the employed Evolutionary Algorithm. a
0.5
Parameter Value
0.45
Population size of first generation 20 Exp. T. Hayashi et al (2005)

Torque coefficient (Cm)

Number of generations 20 0.4 Present Numerical Work
Survival probability 50%
Average probability 33.3% 0.35
Crossover probability 16.7%
Mutation probability 100%
0.3
Mutation magnitude 30%a (i.e., 15%) 0.25
a
This value is multiplied by 0.8 at each generation. For example the mutation
magnitude is only 4% (2%) after 10 generations. Mutation magnitude must be
0.2
decreased during the optimization process to stabilize the population.
0.15
0.1
results. Therefore, the present study relies again on EA. The
0.05
employed parameters are listed in Table 2.
A fully automatic optimization finally takes place, using OPAL 0
(decision-maker for the configurations to investigate), the 0.2 0.4 0.6 0.8 1 1.2 1.4
commercial tool Gambit for geometry and grid generation
(including quality check) and the industrial CFD code ANSYS-Fluent
Speed ratio ( )
to compute the flow field around the Savonius turbine. As a result of b 0.3
the CFD computation the output power coefficient is determined,
and is stored in a result file. The procedure is automated using
journal scripts (Gambit, Fluent) and a master program written in C, 0.25 Exp. T. Hayashi et al (2005)
calling all codes in the right sequence. By checking the values stored
Power coefficient (Cp) Present Numerical Work
in the result file, OPAL is able to decide how to modify the input 0.2
parameters, before starting a new iteration. The fully coupled
optimization procedure is a complex task, which has been
described in detail in previous publications. We thus refer the 0.15
interested reader to see refs. [16,17,19,20] for a complete description
of the procedure. 0.1

info about optimization method

4. Numerical flow simulations 0.05

The optimization procedure can only deliver the right solution, 0

if all evaluations relying on Computational Fluid Dynamics indeed
0.2 0.4 0.6 0.8 1 1.2 1.4
lead to an accurate flow description. From the literature it is known
that an accurate CFD simulation of the flow around a Savonius Speed ratio ( )
turbine is particularly a challenging task, mainly due to its highly
Fig. 5. Validation of computational model: a: torque coefficient, b: power coefficient,
time-dependent nature and to the fact that flow separation plays an both compared to published experimental results for a two-blade Savonius rotor [10].
important role for the efficiency of the system. It is therefore
CFD Solver setup
necessary to check the full CFD procedure with great care. After-
wards, the resulting methodology must be validated.
All flow simulations presented in this work rely on the industrial
software ANSYS-Fluent. The unsteady Reynolds-Averaged
NaviereStokes equations are solved using the SIMPLE (Semi-
Implicit Method for Pressure-linked Equations) algorithm for
pressureevelocity coupling. The flow variables and all turbulent
quantities are discretized in a Finite-Volume formulation using
a second-order upwind scheme. In order to model turbulence, the
realizable ke3 model is employed, which is usually recommended
for rotating bodies. For the present configuration, two-dimensional
simulations are sufficient (no geometry change in the third direc-
tion when excluding boundary effects), so that very fine grids can
be employed.
The unsteady flow is solved by using the Sliding Mesh Model
(SMM). Three complete revolutions are always computed, using an
appropriate, constant value of the time-step; the first revolution is
only used to initiate the correct flow solution, while the flow
properties (in particular the power coefficient Cp and the torque
coefficient Cm) are obtained by averaging the results during the last
two revolutions. It has been checked that the results do not change
noticeably by iterating further in time, so that this present combi-
nation (1 revolution for initializing the flow þ 2 revolutions to
Fig. 4. Grid-independence study for the torque coefficient. compute the target function) has been kept throughout. On
 2622 M.H. Mohamed et al. / Renewable Energy 35 (2010) 2618e2626

a Optimal configuration
-1.017 0.0 -1.017 0.2503

0.182 Without
obstacle

-1.88 -0.88 -1.88 0.133

X1/R Y1/R X2/R Power coeff.

b -1.017 0.0 -1.017 0.2503

Optimal configuration

0.212

0.153
Without
obstacle

-1.88 -0.88 -1.88 0.133

X1/R Y1/R X2/R Power coeff.

Fig. 7. Input parameters of the optimization and power coefficient represented using
parallel coordinates: a: two-blade Savonius turbine; b: three-blade Savonius turbine.
The parameters of the optimal configurations are connected with a thick gray line. The
power coefficient of the conventional turbine is shown with a gray circle. The scales of
both figures have been kept identical to facilitate comparisons.

notice from Fig. 4 that the five coarsest grids are associated with
a large variation of the torque coefficient. On the other hand, all
Fig. 6. Static torque coefficient as a function of the rotor angle q for three different
remaining grids employing more than 71,000 cells lead to a relative
values of Y1 choosing X1/R ¼ 1.4 and X2/R ¼ 1.76: a: two-blade Savonius turbine, b:
three-blade Savonius turbine. variation of the output quantity below 1.8%. Since the cost of a CFD
evaluation obviously increases rapidly with the number of grid
mesh independence test cells, the intermediate grid range between 75,000 and 95,000 cells
has been retained for all further results shown in the present paper.
a standard PC, one evaluation (i.e., three revolutions for one specific Finally, the full CFD model has been validated by comparison
configuration) takes about 170 min of computing time. with published experimental results [10] for a conventional Savo-
After setting up this general CFD model, a grid-independence nius turbine. The results presented in Fig. 5 demonstrate the
study has been carried out for a chosen configuration. Corre- excellent agreement obtained between CFD and experiments for
sponding results are shown in Fig. 4. Several different two- this standard configuration, especially for l > 0.4. Both the torque
dimensional grids of increasing density and quality, composed of
3,400 up to 116,000 cells, have been tested for the two-blade
Savonius turbine with one specified obstacle plate (choosing X1/ Table 4
Optimal configurations.
R ¼ 1.4, Y1/R ¼ 0.24 and X2/R ¼ 1.76 at l ¼ 1.0). It is easy to
Design Parameter Value b Power coeff. Cp
Table 3
Two-blade Savonius turbine X1/R 1.23830 (100.83 ) 0.2503
Acceptable range for the input parameters (parameter space).
Y1/R 0.45390
Parameter Minimum Maximum X2/R 1.09993

X1/R 1.88 1.017 Three-blade Savonius turbine X1/R 1.05632 (80.52 ) 0.2120
Y1/R 0.88 0.0 Y1/R 0.36912
X2/R 1.88 1.017 X2/R 1.38162
 M.H. Mohamed et al. / Renewable Energy 35 (2010) 2618e2626 2623

coefficient and the power output coefficient are found very close to this angle q. Fig. 6 shows the obtained static torque coefficient Cms
the experimentally measured values. As a consequence, the obtained for three different positions of the deflecting plate as
developed CFD procedure can now be used as a reliable evaluation a function of q. The experimental results of refs. [8] and [10] for the
step for the optimization process. conventional turbines are also shown for comparison in Fig. 6. Due
to periodicity, the results are only plotted for q between 0 and 180
5. Results and discussions for the two-blade turbine and q between 0 and 120 for the three-
blade turbine. These computations demonstrate that the obstacle
5.1. Self-starting capability plate has a considerable and globally positive effect on the static
torque coefficient for both configurations, in particular for the two-
One important issue associated with wind energy is the self- blade design. The conventional two-blade turbine shows a very
starting capability of the system. For decentral, low-cost applica- large variation of the static torque coefficient as a function of q, with
tions as considered here, it is essential to obtain a self-starting negative values around q ¼ 140 e170 (no self-starting). For all
system. To investigate this issue, the static torque exerted on investigated positions involving an obstacle, the negative torque
a turbine at a fixed angle has been computed by CFD as a function of region completely disappears, with a minimum value of Cms higher

Fig. 8. Comparison of the instantaneous flow structures when the advancing blade is in vertical position for the optimal configurations at l ¼ 0.7: a: two-blade turbine; b: three-
blade turbine. Zoom on the vicinity of the turbine (full CFD domain is much larger).
2624 M.H. Mohamed et al. / Renewable Energy 35 (2010) 2618e2626

than 0.07. Apart from that, the evolution as a function of q is similar R ¼ 1.177 (Fig. 3). With these three factors, the position of the
to that of the conventional turbine. For the three-blade case, the deflecting plate is perfectly determined, and the angle b can be
obstacle plate improves the self-starting capacity for part of the q- deduced as well.
range. As a whole, employing an obstacle plate improves noticeably In a previous study considering only the two-blade Savonius
the self-starting properties for the two-blade configuration and (to turbine [11], a small range of variation had been defined for the
a lesser extent) for the three-blade configuration. A self-starting three degrees of freedom (so-called parameter space). Here, a much
capability ðCms > 0Þ is always obtained in principle at any angle, larger accessible domain has been prescribed, as documented in
which is a major advantage. Table 3. In this manner, a truly optimal solution can be obtained
considering a large set of possible parameters. Of course, when
choosing the parameter space, it must be checked that the shield-
5.2. Optimization of the obstacle position ing obstacle cannot come into direct contact with the rotor
(kX1k > R and kX2k > R). The corresponding positions of the
Finally, the mathematical optimization procedure described obstacle all lead to configurations that shield partially the returning
previously (Evolutionary Algorithms relying on automated evalu- blade.
ations through CFD) can be employed to find the optimal position of Finally, the optimization process thus involves simultaneously
the obstacle. This is done for a speed ratio l ¼ 0.7, considering three parameters (or degrees of freedom): X1, Y1 and X2. For each
a fixed incident wind velocity U ¼ 10 m/s. This value of l is retained, geometrical configuration one single objective (power output
since it is known from the literature that it corresponds to the zone coefficient) is determined by CFD evaluations, and should be
of peak power coefficient of the conventional Savonius turbine. maximized by the optimization procedure.
As explained previously, three degrees of freedom are left In Fig. 7 a parallel coordinate representation has been chosen
simultaneously to the OPAL optimizer: X1 and Y1 define the upper since it is the most popular way to analyze output data from
tip of the shielding obstacle; the value X2 is then sufficient to define optimization involving several degrees of freedom. Such figures
the position of the lower tip, since Y2 is taken constant, with Y2/ might be at first difficult to understand. Each thin connecting line

Savonius without obstacle

Three blade Savonius
Savonius with obstacle without obstacle
a (optimal configuration) a Three blade with obstacle
0.5
% Relative increase 0.5 (optimum configuration)
0.45 % Relative increase
Torque coefficient (Cm)

Torque Coefficient (Cm)

0.4
0.4
0.35

0.3
0.3
0.25

0.2
0.2
0.15

0.1
0.1
0.05

0
0
0.2 0.4 0.6 0.8 1 1.2 1.4
0.2 0.4 0.6 0.8 1 1.2
Speed ratio (λ)
Speed ratio (λ)
b b
0.3 100 % 0.25 100
Power Coefficient (Cp)
Power coefficient (Cp)

0.25
Relative incease (%)

Relative incease (%)

80 0.2 80

0.2
60 0.15 60

0.15

40 0.1 40
0.1

20 0.05 20
0.05

0 0 0 0

0.2 0.4 0.6 0.8 1 1.2 1.4 0.2 0.4 0.6 0.8 1 1.2
Speed ratio (λ) Speed ratio(λ)
Fig. 9. Performance of the optimized configuration (filled squares) compared to the Fig. 10. Performance of the optimized configuration (filled squares) compared to the
conventional two-blade Savonius turbine without obstacle (empty squares): a: torque conventional three-blade Savonius turbine without obstacle (empty squares): a: torque
coefficient; b: power coefficient. The corresponding relative increase is shown with coefficient; b: power coefficient. The corresponding relative increase is shown with
stars. stars.
 M.H. Mohamed et al. / Renewable Energy 35 (2010) 2618e2626 2625

represents all numerical parameters associated with one specific 6. Conclusions

configuration, where each parameter is associated with its own
vertical axis. The first three columns therefore show the value of The Savonius turbine is a promising concept for small-scale
the three free parameters, X1/R, X2/R and Y1/R; the last column on wind-energy systems, but suffers from a poor efficiency. Therefore,
the right corresponds to the value of the objective function, here the major objective of the present study is to identify an improved
the power coefficient Cp. The scale of the first parameter (X1/R) is for design, leading to higher values of the power coefficient and of the
instance bounded between 1.88 (minimum) and 1.017 static torque of the Savonius turbine, thus obtaining a higher effi-
(maximum). By following a single line, the reader can therefore ciency and better self-starting capability. For this purpose, an
determine quantitatively the values of all parameters associated obstacle plate is introduced that shields partly the returning blade
with one configuration. The optimal solution is shown with a thick of a Savonius turbine (using either two or three blades) and that
gray line. More details concerning parallel coordinates can be found optimizes the wind direction toward the advancing blade. The flow
for instance in ref. [21]. solution is obtained by accurate CFD simulations. A validation
The results presented in Fig. 7 indicate that the considered against published experimental measurements shows that the
objective is indeed considerably influenced by the three free employed numerical procedure relying on the realizable kee
parameters, X1, Y1 and X2. As a whole, 210 different geometrical turbulence model can be employed for a quantitative analysis of the
settings have been evaluated by CFD, requesting 22 days (respec- performance, provided a sufficiently fine grid is used.
tively 24 days) of computing time on a standard PC for the three- The installation of the obstacle improves the self-starting
blade (resp. the two-blade) Savonius turbine. Note that the user- capability for almost all configurations. While the conventional
waiting time could be considerably reduced by carrying out the two-blade Savonius turbine shows negative values for the static
requested CFD in parallel on a PC cluster [16]. Such a parallel moment in some range, the obstacle leads to a positive static
procedure, already implemented in OPAL, has not been used in the moment value at any angle for both designs.
present case but could reduce the needed time by more than an Finally, the position of the shielding obstacle has been optimized
order of magnitude, as demonstrated in other studies. in a fully automatic manner, in order to obtain the best possible
The optimal configurations (highest point in the right column in performance. The optimization relies on Evolutionary Algorithms,
Fig. 7, all corresponding parameters being connected by a thick gray while all geometrical configurations are evaluated in an automatic
line) can now readily be identified for l ¼ 0.7. The optimum manner by CFD. This optimization procedure is able to identify
obstacle position, optimum angle b and corresponding optimal considerably better configurations than the original Savonius
power coefficient Cp are listed in Table 4. turbines, leading in particular to a relative increase of the power
When compared with the standard Savonius turbines (without output coefficient by 27.3% (resp. 27.5%) for the two-blade (resp. the
shielding obstacle), the optimal point found by the optimization three-blade) Savonius rotor. This gain is also observed for the tor-
procedure corresponds to an absolute increase of the power coef- que coefficient, and is even higher for larger values of l. Therefore,
ficient by 0.068 (respectively 0.058) at l ¼ 0.7 for the two-blade the overall effect of this obstacle is extremely positive for both
(resp. for the three-blade) Savonius turbine. This means a relative designs. Considering the obtained output power coefficient and the
increase of the performance (measured by the power output cost and complexity of the rotor, the two-blade configuration is
coefficient) by 27.3% for the two-blade Savonius turbine, and by clearly better than the three-blade turbine. This optimal configu-
27.5% for the three-blade case, compared to the conventional ration leads to a peak power output coefficient of 0.258 at l ¼ 0.8,
Savonius design without any obstacle. and seems therefore very promising for wind energy generation in
Note that we have been at first surprised to obtain an optimal urban areas.
geometry leading to b > 90 for the two-blade turbine. After From the technical point of view most wind turbines have
a thorough flow analysis, partially shown in Fig. 8, it is possible to upwind rotors and many existing systems rely on a tail vane for
understand finally that the flow direction induced by the obstacle optimal alignment into the wind direction. A similar technical
at b ¼ 100.83 is indeed optimal for the advancing blade in the solution should be used for the Savonius turbine using an obstacle
relative reference frame. Furthermore, for most configurations shielding the returning blade. In this manner the orientation of the
associated with b < 90 , the flow behind the shielding obstacle system can be simply, efficiently and automatically controlled. As
points partly toward the returning blade and thus increases the a whole, the optimized configuration with the shielding obstacle is
reverse moment; this effect is reduced for the optimal configura- only slightly more complex, more expensive and heavier than the
tion. Due to the different interaction process between the blades, original system. Therefore, the improved power and torque coef-
the situation is different for the three-blade turbine, for which ficients should easily compensate these drawbacks within a short
b ¼ 80.52 indeed leads to optimal flow conditions. These different time after installation.
values of b are a direct consequence of the different trade-off
between the positive and reverse moments induced by the
Acknowledgments
advancing (resp. returning) blade(s) in the two cases. The optimal
geometry of the shielding obstacle is therefore highly dependent on
The Ph.D. work of Mr. Mohamed is supported financially by
the specific rotor configuration.
a bursary of the Egyptian government. Interesting discussions with
It is now important to check how this gain would change as
G. Warschau are gratefully acknowledged.
a function of l, since such a turbine must be able to work also for
off-design conditions. Therefore, the performance of the optimal
configurations has been finally computed for the full range of useful References
l-values, as shown in Figs. 9 and 10. These figures demonstrate that
[1] Gupta R, Biswas A, Sharma KK. Comparative study of a three-bucket Savonius
the improvement of both torque coefficient and power output
rotor with a combined three-bucket Savonius-three-bladed Darrieus rotor.
coefficient is observed throughout for all values of l, compared to Renewable Energy 2008;33(9):1974e81.
the conventional Savonius turbine without obstacle. The absolute [2] Menet J, Valdés L, Ménart B. A comparative calculation of the wind turbines
gain for Cp increases even slightly with l at first, the relative capacities on the basis of the L e s criterion. Renewable Energy 2001;22
(4):491e506.
increase being highest for the largest values of l considered in the [3] Menet J. A double-step Savonius rotor for local production of electricity:
present study. a design study. Renewable Energy 2004;29(11):1843e62.
2626 M.H. Mohamed et al. / Renewable Energy 35 (2010) 2618e2626

[4] Murai Y, Nakada T, Suzuki T, Yamamoto F. Particle tracking velocimetry [13] Van den Braembussche RA. Numerical optimization for advanced turboma-
applied to estimate the pressure field around a Savonius turbine. Measure- chinery design. In: Optimization and computational fluid dynamics. Berlin,
ment Science and Technology 2007;18:2491e503. Heidelberg: Springer-Verlag; 2008. p. 147e88.
[5] Kamoji MA, Kedare SB, Prabhu SV. Performance tests on helical Savonius [14] Van den Braembussche RA, Brouckaert JF, Paniagua G, Briottet L. Design and
rotors. Renewable Energy 2009;34(3):521e9. optimization of a multistage turbine for helium cooled reactor. Nuclear
[6] Huda MD, Selim MA, Islam AKMS, Islam MQ. Performance of an S-shaped Engineering and Design 2008;238(11):3136e44.
Savonius rotor with a deflecting plate. RERIC International Energy Journal [15] Veress Á, Van den Braembussche RA. Inverse design and optimization of
1992;14(1):25e32. a return channel for a multistage centrifugal compressor. Journal of Fluids
[7] Saha UK, Rajkumar MJ. On the performance analysis of Savonius rotor with Engineering, Transactions of the ASME 2004;126(5):799e806.
twisted blades. Renewable Energy 2006;31(11):1776e88. [16] Thévenin D, Janiga G. Optimization and computational fluid dynamics. Berlin,
[8] Irabu K, Roy JN. Characteristics of wind power on Savonius rotor using Heidelberg: Springer-Verlag; 2008.
a guide-box tunnel. Experimental Thermal and Fluid Science 2007;32 [17] Hilbert R, Janiga G, Baron R, Thévenin D. Multi-objective shape opti-
(2):580e6. mization of the flow around a heat exchanger using parallel genetic
[9] Menet J. Aerodynamic behaviour of a new type of slow-running VAWT. In: algorithms. International Journal of Heat and Mass Transfer 2006;49:
Wind energy proceedings of the Euromech Colloquium, vol. 43. Springer; 2567e77.
2007. 235e240. [18] Mohamed MH, Janiga G, Thévenin D. Performance optimization of a modified
[10] Hayashi T, Li Y, Hara Y. Wind tunnel tests on a different phase three-stage Wells turbine using non-symmetric airfoil blades. In: ASME Turbo Expo
Savonius rotor. JSME International Journal, Series B: Fluids and Thermal Conference, (GT2008-50815). Berlin, Germany; 2008.
Engineering 2005;48(1):9e16. [19] Thévenin D, Zähringer K, Janiga G. Automatic optimization of two-dimen-
[11] Mohamed MH, Janiga G, Pap E, Thévenin D. Optimal performance of a Savo- sional burners. European Combustion Meeting. ECM05, Louvain-La-Neuve.
nius turbine using an obstacle shielding the returning blade. In: Ninth inter- Belgium; 2005. p. 240/1e6.
national congress of fluid dynamics and propulsion. ASME-ICFDP9, (ICFDP9- [20] Janiga G, Thévenin D. Reducing the CO emissions in a laminar burner using
EG-249). Alexandria, Egypt; 2008. different numerical optimization methods. Journal of Power and Energy
[12] Menet J, Nachida B. Increase in the Savonius rotors efficiency via a parametric 2007;221(5):647e55.
investigation. European wind energy conference EWEA e poster presenta- [21] Ilsenberg A. Visual data mining with Parallel Coordinates. Computational
tions. Aerodynamics and Aeroacoustics; 2004. Statistics 1998;13(1):47e63.