Renewable Energy 179 (2021) 1345e1359

Contents lists available at ScienceDirect

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

An OpenFOAM based study of Savonius turbine arrays in tunnels for

power maximisation
Rajagopal Vinod Bethi a, Santanu Mitra b, *, Pankaj Kumar c
a
Simulation Engineer, UFI Innovation Center, Aerocity, New Delhi, 110037, India
b
Department of Mechanical Engineering, Shiv Nadar University, Gautam Budh Nagar, Uttar Pradesh, 201314, India
c
Former Scientist, Institute of High Performance Computing, A*STAR, Singapore

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

Article history: There has been a significant disruption in the energy market because of the pandemic. Now the focus is
Received 20 January 2021 shifting towards investing more in renewable energy to address the present health and energy crisis.
Received in revised form Wind power has gained special prominence in our clean energy transition as wind energy is the easiest
31 May 2021
to harvest, most efficient resources, and lowest carbon emitter among the available technologies. The
Accepted 14 July 2021
Available online 21 July 2021
prime focus of this study is towards implementing an array configuration of Savonius turbines beside the
train track to maximize the power production. A model is set up on OpenFOAM platform and studied for
the different arrangements of turbine clusters. The effect of energy production by the series of turbines
Keywords:
Wind energy harvesting
was assessed by varying distance among turbines in a staggered or inline manner and the location from
Array of turbines the train as well. A systematic characterization of the optimum Savonius turbine cluster in a railway
OpenFOAM tunnel, which is relatively an unexplored area in the energy harvesting community, has been taken up in
Computational fluid dynamics this study. It is exciting to note that our computer-oriented result matches well with the benchmark
solution and proved to be a viable green energy resource that can possibly erase the energy crisis at
remote places.
© 2021 Elsevier Ltd. All rights reserved.

1. Introduction As a potential technology, Savonius Vertical Axis Wind Turbnes

(VAWT) can be placed inside the tunnels to harvest the energy
Generating power from unconventional renewable energy re- produced by the trains. The authors have already showcased that
sources has become necessary as the demand for power is with design modifications it is possible to harvest wind energy
increasing but the non renewable energy resources that we are from moving trains inside tunnels using Savonius turbines [2]. In
currently reliant on are depleting fast and are also harmful to the the study, the traditional Savonius turbine design failed to harvest
environment. Rapid electrification is taking place in every sector energy under turbulent flow but with the modified turbine, the
and coal burning or fuel based power production is slowly authors exhibited the possibility of harvesting wind energy from
becoming secondary by the advent of renewable energy resources. such an unconventional source. It was also shown that the drag
Power generation from renewable energy resources will not be induced on the train by the turbine is negligible compared to the
sufficient for future demands and as per the 2016 data on “share of energy generated by the turbines and that the turbine can harvest
energy from renewable resources in gross final consumption of energy even when the trains are moving at very low speeds. The
energy” in the EU, shows that only 29.6% of the energy is obtained placement of the turbine with respect to the train is a major factor
from renewable resources [1]. Hence new avenues of power gen- and it was found that the turbines have to be placed as far away
eration has to be explored where little to no progress has been from the train and tunnel walls as possible to avoid interference
made to harvest energy from unconventional resource. One such from boundary layer phenomena. The average maximum energy
area is the harvesting of energy from moving vehicles in tunnels harvested was 738.92W when a turbine of 0.7 m diameter was
where trains travel at high speeds producing vast amounts of wind placed at a distance of 0.5 m from the train. Consequently, in this
energy. study, a concept of array of turbines in tunnel in introduced to in-
crease the total power generated. There is a general lack of litera-
* Corresponding author. ture that directly addresses the arrangement and placement of
E-mail address: [email protected] (S. Mitra).

https://doi.org/10.1016/j.renene.2021.07.071
0960-1481/© 2021 Elsevier Ltd. All rights reserved.
R.V. Bethi, S. Mitra and P. Kumar Renewable Energy 179 (2021) 1345e1359

Savonius VAWTs in such scenarios. It is crucial to understand the OpenFOAM solvers are based on Finite Volume Method (FVM)
effects of placing an array besides the tunnel to maximize the po- and the fluid flow is governed by the Navier-Stokes equations. The
wer generation. However, there has been many studies on wind velocity components and pressure fields of the fluid flow that
farms and how to effectively place each turbine with respect to the fluctuates with respect to both time and space, known as turbu-
other in open environment. The present scenario does not have lence, are averaged to form Reynolds Averaged Navier-Stokes
enough leeway in terms of placement of turbines as the space is (RANS) equations.
restricted between the train and the tunnel walls. The wakes pro-
duced from the preceding turbines affects the performance and vUi
¼0 (4)
power generation [3]. The 1st turbine receives relatively less tur- vx
bulent flow and hence would be able to generate more power than
the other turbines [4]. The farther the turbines are spaced apart the vðrUi Þ vðrUi Uj Þ ́ ́
þ vP þ vð2 mS  ru u Þ
¼ vx
vt vxj ij i j
lesser is the effect of the wakes of the preceding turbine but after a i
(5)
certain distance the negative effect does not vary much [4]. The vxj
wind farm layout has to be optimized for every scenario as one ́ ́
layout does not perform well under various conditions [5]. The The term ui uj in the RANS equations is called as Reynolds stress
effect of placing multiple turbines can also significantly affects the tensor which is an unknown and needs to be evaluated using tur-
flow thereby completely changing the dynamics of the system [6]. bulence modelling such as kε, ku or ku  SST.
The wake can also sometimes enhance the turbine performance in The pimpleDyMFoam RANS solver was used because of its dy-
certain locations while inhibiting in other configurations [7]. The namic mesh handling capabilities. The PIMPLE algorithm is a
wake propagation can interact with other elements in its path to combination of PISO and SIMPLE algorithms. The kε turbulence
the next turbine and can significantly be different if there are no model was used to solve the RANS equations [9]. The gradients of
interference in the path between two consecutive turbines [8]. pressure and velocity are determined using the default numerical
Considering the above points, the authors felt that it was scheme, Gauss linear. The convective terms and diffusion terms are
necessary to investigate the performance of an array of turbines for determined by a second order upwind scheme and two corrector
harvesting wind energy from the trains that are moving inside loops are employed with a maximum of 100 iterations per loop. The
tunnels. To that end, a numerical study is set up to study and dynamic mesh motion is handled by sixDoFRigidBodyMotion
optimise various array configuration that can be utilised inside the solver which is a sub-class of dynamicMotionSolverFvMesh in
tunnels. Due to space constraint only the 0.5 m diameter turbine is OpenFOAM [10]. The time step for the transient simulations are
considered. The previously developed model on OpenFOAM is adaptively changed to maintain the Courant number below 0.1 in
utilised with minor modifications to accommodate multiple order to ensure stability and accuracy of the solutions.
rotating objects.
2.2. Geometry

2. Methodology Two-dimensional Savonius rotor and train geometry (see Fig. 1)

are modelled based on the circle line of the Singaport Mass Rapid
2.1. Governing equations and numerical models Transport (SMRT) as shown in Fig. 2 which is also similar to our
previous study [2]. The rotor consists of two semi-circular blades of
The performance of a turbine largely depends on two parame- diameter d ¼ 0.2m, blade thickness t ¼ 0.025m and overall rotor
ters, the coefficient of power (CP) and the coefficient of torque (CT). diameter D ¼ 0.5m. The width of the train geometry is 1.6m which
The coefficient of power of a turbine is an indication of the power is in accordance to train model of Alstom Metropolis C830C [11]
generated by the turbine from a theoretically calculated maximum that is used in the circle line. The length of train varies based on the
energy that is possible to harvest. The coefficient of torque in a array employed. The width of the domain is 2.9m which is in
similar sense is the ratio of torque generated by the turbine and accordance to the tunnel dimensions [12]. The geometry is sepa-
theoretically calculated maximum torque that it can generate. The rated in to two domains, inner rotating circular domain where the
two parameters are given by turbine is placed and rectangular outer domain with the train.
Coupling between the two domains is achieved by Arbitrary Mesh
P Interface (AMI) using cyclicAMI boundary condition utility in
CP ¼ (1)
0:5rAU 3∞ OpenFOAM. AMI allows transferring of field variables between the
stationary domain patch and the rotating domain patch. The
T contribution from overlapping faces are defined by weights which
CT ¼ (2) should be equal to 1 [13]. The train, tunnel and turbine surfaces are
0:5rAU 2∞ R
treated as walls with no-slip boundary condition. At the inlet the
where r is the air density (1.225Kgm3), A is the area of the turbine velocity of air is specified that is equal to the velocity of the train.
(m2), U is the free stream velocity of air (m/s), R is the radius of the The outlet is given the $internalField boundary condition in
turbine (m), T is the dynamic torque and P is the power produced by OpenFOAM which virtually is similar to having an infinitely long
turbine given by domain. Symmetry is considered along the center line of the train
and tunnel. The velocity components inside the domain are set to
P ¼ Tu (3) zero at the start of the simulation. The turbine is at rest initially and
is allowed to rotate along its vertical axis during the simulation run
where u is the angular velocity of the turbine (rad/s). The Open- time. The actual length of the domain is changed based on the
FOAM forces library is used to monitor the torque produced and the number of turbines to be simulated and the distance between
mesh motion solver writes the angular velocity of the turbine at them. This is done to reduce number of mesh elements and thereby
every timestep. significant reducing computational requirements.

1346
R.V. Bethi, S. Mitra and P. Kumar Renewable Energy 179 (2021) 1345e1359

Fig. 1. 3D representation of the proposed design.

Fig. 2. Plan view of the proposed design with boundary conditions.

2.3. Mesh 2.4. Validation

The meshing techniques used is also exactly similar to our In our previous study, the numerical simulation was verified
previous study [2]. To satisfy wall function requirements, necessary with the experimental setup mentioned in Ref. [14]. Fig. 4 shows
when turbulence model is used, a yþ value of 149 was set and the comparison between coefficients of power as a function of
achieved. Mesh refinement is concentrated around the turbine and various tip speed ratio as performed in the experimental study. In
train walls as shown in Fig. 3, similar to Bethi et al. [2]. The total Fig. 4 the numerical study shows a good agreement with the
number hexahedral mesh elements are in the range of 4  105 to experimental study as the trend and values of both the plots match
12  105. Residual RMS error value around 1  106 has been set very well. The slight difference between the numerical and exper-
and achieved in all the simulations which states that the simulation imental results may be due to the blade tip effects of the turbine.
is well converged. The numerical model does not account for blade tip effects. Though
there is only one cell in the third direction, OpenFOAM calculates

Fig. 3. Mesh used for the simulations.

1347
R.V. Bethi, S. Mitra and P. Kumar Renewable Energy 179 (2021) 1345e1359

placed at 10m and 12.5m have almost similar values throughout. It

is observed that when the turbines are placed any closer than 10m
there is a considerable drop in performance and beyond 10m the
difference in performance in minimal. Table 1 contains the power
produced by the turbines placed at different distances. From the
values of power generated, an distance of 10m is selected for sub-
sequent simulations as the power generated becomes saturated
beyond 10m (as seen for a distance of 12.5m).

3.2. Performance evaluation of 1st turbine

The total power produced by an array of turbine depends on the

Fig. 4. Tip Speed Ratio versus CP; comparing experimental results from Ref. [14] and
CFD results.
cumulative performance of all the turbines in the array. Hence, it is
essential to study the performance of the turbines under different
configurations. Bethi et al. showed that the power generation of the
the surface area based on the dimensions of that one cell in the turbine depends on the placement of the turbine with respect to
third direction. Hence, the model is dimensionally same to that of the train and tunnel walls [2]. In this study, it was shown that the
the experimental setup and it is agreeable to use standard RAS kε farther the turbine is placed from both the train and tunnel walls,
model as the turbulence solver for the simulations. the more power it was able to generate. The study investigated
performance of the turbines for distances of 0.1m, 0.25m, 0.5m and
0.75m from the train and the turbine placed at a distance of 0.5m
3. Results and discussions
from the train generated the most power out of all the cases.
However, in this study we have chosen 0.3m, 0.5m and 0.7m as the
3.1. Optimizing distance between subsequent turbines
distances to reduce the effects of the train and tunnel walls on the
performance of the turbine.
The flow past a turbine is fraught with wake vortices or wake
To that end, a turbine of 0.5m was placed at different distances
turbulence [15]. The performance of trailing turbines depend on
from the train and the performance of the turbines were studied.
the wake produced by the leading turbine. To that end, blade
Fig. 8 shows the variation in CT for the turbines placed at different
optimization of Savonius VAWT to reduce the extent of vortex
distances with respect to train. The turbine placed far away at 0.7m
shedding has received considerable attention in literature [16,17].
exhibits slow starting characteristics which depicts that the flow
The present study however is focused on optimizing the placement
speed is lower in that region. Fig. 9 shows the variation in angular
of Savonius VAWTs in an array to maximize power generation from
velocities. Evidently, the turbine placed at a distance of 0.5m per-
high velocity winds produced by SMRT trains travelling inside
forms better than the other two cases.
tunnels. The space available inside the tunnel between the train
The Table 2 details the power produced by the turbines for a
and tunnel walls is minimal and hence only a single turbine is
period of 4 s. It is observed that the turbine placed at 0.5m distance
placed in the horizontal direction. Subsequent turbines are placed
from train can harvest more power. The power generated by the
in an array in the vertical direction. It is to be noted that the tur-
these turbine would be much higher than the subsequent turbines
bines in the array are geometrically identical.
that are placed in an array.
Initially, two turbines of 0.5 m diameter are placed inline with
respect to each other at distance of 0.5 m from the train. Fig. 5
shows a screenshot from the array simulation. The distance be- 3.3. Performance evaluation of 2nd turbine
tween the turbines is varied and the effect of the wake produced by
the first turbine on the second turbine is studied. It is to be noted After investigating the performance of the 1st turbine, the case
that both the turbines start from u ¼ 0 and are studied for 4 s (the of an array of turbines is set-up in OpenFOAM. Currently, the array
time taken for the train to pass by the turbines). Fig. 6 show the contains only two turbines, so as to understand the effects of the 1st
variation of CT with time for the second turbine in the array placed turbine on the performance of the 2nd turbine. The placement of
at a distance of 5m, 7.5m, 10m and 12.5m from the first turbine. The the 2nd turbine, with respect to train, is varied for a given place-
distances are calculated from the center of rotation of the first ment of the 1st turbine. It is to be noted that both the turbines start
turbine to the center of rotation of the second turbine. from u ¼ 0 at t ¼ 0 and the simulations are run until t ¼ 4.
From Fig. 6 it is observed that the as the distance between the
turbine is increased, the characteristics of CT curve is better i.e the 3.3.1. When 1st turbine is placed at 0.3m
peak values are higher and production of negative torque and The 2nd turbine is placed at different distances from the train
fluctuations in torque production is lesser. Fig. 7 shows the varia- behind the first turbine and the power generated is compared.
tion of u with time for the turbines. It is evident that the turbines Fig. 10 shows the CT variation for the different cases considered. The

Fig. 5. Screenshot from array simulation.

1348
R.V. Bethi, S. Mitra and P. Kumar Renewable Energy 179 (2021) 1345e1359

Fig. 6. CT of 2nd turbine placed at distance of (a) 5m, (b) 7.5m, (c) 10m and (d) 12.5m from the 1st turbine.

Fig. 7. u of 2nd turbine placed at distance of (a) 5m, (b) 7.5m, (c) 10m and (d) 12.5m from the 1st turbine.

Table 1 From the simulations it is observed that some of the vortices

Power generated by 2nd turbine placed at distance of 5m, 7.5m, 10m and 12.5m from shed from the 1st turbine strikes back and forth between the tunnel
the 1st turbine.
wall and the train wall as it traverses down stream to the 2nd
Distance from Distance from Power generated turbine. The vortices breakdown after each collision with the walls.
train 1st turbine The remaining vortices travel in-line to the 2nd turbine. Therefore,
0.50 m 5m 162.54 W the turbine placed close to the tunnel wall (at 0.7m from the train),
7.5 m 175.57 W exhibits better performance.
10 m 189.60 W
12.5 m 191.44 W

3.3.2. When 1st turbine is placed at 0.5m

turbine placed at 0.7m from the train exhibits higher CT values than The same method as before is followed and the 2nd turbine is
that of the other cases. The speed of the turbine is also higher than placed at different distances from the train behind the first turbine.
the other cases as shown in Fig. 11. The 1st turbine is placed at 0.5m from the train and the power
The Table 3 details the power generated by the turbines for generated by the 2nd turbine is compared. Fig. 12 shows the CT
different cases. The turbine placed at 0.3m generates the least variation for the different cases considered. The turbine placed at
amount of power. The power loss is the difference between the 0.7m from the train exhibits better CT values than that of the other
power produced by the 2nd turbine and the 1st turbine and is given cases. The speed of the turbine is also higher compared to the other
by the formula cases as shown in Fig. 13. This trend is similar to the previous case
where the staggered arrangement of the turbine performed better.
 
P1st Turbine  P2nd Turbine The propagation of vortices are affected by the train and tunnel
Ploss ¼ 100 (6)
P1st Turbine walls and hence the turbine placed in-line at 0.5m produces
reasonable power as it is relatively affected less by the those
1349
R.V. Bethi, S. Mitra and P. Kumar Renewable Energy 179 (2021) 1345e1359

Fig. 8. CT of turbines placed at a) 0.3 m b) 0.5m c) 0.7m.

Fig. 9. u of turbines placed at a) 0.3 m b) 0.5m c) 0.7m.

Table 2 Table 5 shows the power generated and power loss. The power
Power generated by 1st turbine at different distances. loss is lesser because the 1st turbine generates least power out of all
possible scenarios.
Diameter Distance from train Power generated
From the above simulations it is observed that the staggered
0.50 m 0.3 m 379.76 W arrangement of turbines are able to generate more power. The
0.5 m 464.78 W
0.7 m 409.16 W
farther the horizontal space between the turbines, the better the
performance of the 2nd turbine. The combination of placing the 1st
turbine at 0.3m from the train and the 2nd turbine at 0.7m from the
train produces the most power. Among all the possible in-line ar-
vortices. The turbine placed at 0.3m generally performs poor due to
rangements, the turbines placed at 0.5m produces the most power.
close proximity with the train walls. The flow near train walls
The in-line arrangements of 0.3m and 0.7m generates less power
contains the vortices shed from the train as well as the vortices
due to their close proximity to wall boundaries. Therefore, by
shed from the turbine.
placing the turbine in staggered arrangement (1st turbine at 0.3m
The Table 4 details the power generated by turbines of different
and 2nd turbine at 0.7m) we could harvest maximum energy.
cases. The power loss of the 2nd turbine is also relatively higher due
Fig. 16 gives a surface representation of the power generated for all
to the power generated by the 1st turbine being the highest among
possible combinations interpolated from the simulated data.
all the three cases considered.
It should be noted that the data holds valid only for a 2 turbine
array arrangements. Moreover, it was assumed that both the tur-
3.3.3. When 1st turbine is placed at 0.7m bines start from u ¼ 0 at t ¼ 0 s which is not the case for a realistic
Similarly the same procedure is followed for when the 1st tur- scenario. The 2nd turbine starts to rotate and harvests energy
bine is placed at 0.7m distance from the train. It is evident from without any interference from the 1st turbine until the train rea-
Figs. 14 and 15 that the trend is the same as the previous case with ches the 1st turbine. Since the turbines are placed at distance of
the turbine at 0.3m performing much better than the other cases. 10m from each other, it would take t z 0.555 s for the train to travel
The turbine placed at 0.7m performs very poorly. from one turbine to the other when travelling at 18 m/s. Such a

1350
R.V. Bethi, S. Mitra and P. Kumar Renewable Energy 179 (2021) 1345e1359

Fig. 10. CT of turbines placed at a) 0.3 m b) 0.5m c) 0.7m.

Fig. 11. u of turbines placed at a) 0.3 m b) 0.5m c) 0.7m.

Table 3 considered consists of four turbines rather than two turbines to

Power generated by 2nd turbine when placed behind 1st turbine at 0.3m
investigate the effects of power loss across multiple turbines.
Placement of 1st Placement of 2nd Power generated Power loss Initially, a staggered arrangement where the turbines are alter-
Turbine w.r.t Train Turbine w.r.t Train natingly placed at distance of 0.3m and 0.7m from the train as
0.30 m 0.3 m 146.49 W 61.43% shown in Fig. 17 is considered. This arrangement generated the
0.5 m 225.08 W 40.73% most power in a two turbine array. The variation of CT is shown in
0.7 m 274.04 W 27.84% Fig. 18. It is to be noted that the simulation time is different from the
time denoted here. Here, t ¼ 0 when u ¼ 0 and starts to advance
when the turbine starts to rotate. Hence, the time for each turbine
simulation would be more accurate in prediction of the total power starts at different instances with respect to the simulation time. The
generated. Hence, to that end, the study also investigates the case of total time a turbine rotates is kept constant at t ¼ 4 s (can be
dynamic starting of turbines. In such a simulation, the trailing interpreted that at t ¼ 4 s the train passes by the turbine
turbine starts to harvest energy free of any influences from pre- completely) for comparison purposes.
ceding turbines. The trailing turbine then experiences the wake of There is marked difference in performance between the 1st
the preceding turbine starting from rest as soon as the train reaches turbine and the rest, which was expected, but the difference in
it. Thus at t ¼ 0.555 s, the 2nd turbine (trailing turbine) rotates at performance between the last 2 turbines is less; see Fig. 19. There
certain rpm achieved from wind generated by the train alone and seems to be a huge drop in performance between the 2nd turbine
the 1st turbine starts to rotate from u ¼ 0. and the 3rd turbine. Though the difference between the 3rd and 4th
turbine is minimal, the total power generated by the array is very
3.4. Series of turbines less. Table 6 details the power generated by each turbine in the
array. Here, the power loss for each turbine is calculated with
3.4.1. Case study I: 0.3me0.7m staggered array respect to the 1st turbine. The 3rd turbine generates less power
An OpenFOAM model is set-up to account for the non- than the 4th turbine since the 3rd turbine is placed near the train
simultaneous dynamic starting of the turbines. The array wall and as mentioned in previous section, the 3rd turbine has to

1351
R.V. Bethi, S. Mitra and P. Kumar Renewable Energy 179 (2021) 1345e1359

Fig. 12. CT of turbines placed at a) 0.3 m b) 0.5m c) 0.7m.

Fig. 13. u of turbines placed at a) 0.3 m b) 0.5m c) 0.7m.

Table 4 Ideally, we could specifically design each and every turbine in the
Power generated by 2nd turbine when placed behind 1st turbine at 0.5m array for maximum power generation. The farther down the array a
turbine is placed, more modifications will be necessary to improve
Placement of 1st Placement of 2nd Power generated Power loss
Turbine w.r.t Train Turbine w.r.t Train its performance. This would become expensive and hence simpler
methods of improvements are pursued here. One such method is to
0.50 m 0.3 m 171.30 W 63.14%
0.5 m 189.60 W 59.21%
change the direction of rotation of every other turbine. Therefore,
0.7 m 209.30 W 54.97% the 1st and the 3rd turbines rotates in a counter-clockwise direc-
tion, whereas the 2nd and the 4th turbines rotates in the clockwise
direction as shown in the Fig. 20. This would ensure that the wake
harvest energy under the influence of both, the train vortices and patterns are disturbed and in some sense cancel out the effects of
the preceding turbine vortices as well. The total power generated vortices produced by two turbines rotating in opposite direction.
by the staggered arrangement is 846.71W. Fig. 21 show the variation of CT for the turbines under such a
The performance drop between the 2nd turbine and the 3rd configuration.
turbine was unexpected. This suggests that the staggered config- The starting phase for the counter-clockwise rotating turbine
uration with more than two turbines is not adept enough to shows positive peaks and the starting phase for the clockwise
maximize the power generation from the SMRT trains. This leads us rotating turbines shows negative peaks. This is due to sign
to investigate methods to improve the performance of the turbines convention and it is assumed that torque is positive in counter-
close to the tail of the array. clockwise direction. The magnitude of the torque is to be
compared and it shows clear improvements over the previous
staggered arrangement. The performance of the 3rd and 4th tur-
3.4.2. Case study II: 0.3me0.7m staggered array with counter bine is seen to improve; see Fig. 22. Here, the magnitude of u is
rotating turbines shown and hence all the values are positive.
Methods to improve the power generated from the staggered Table 7 details the power generated by each turbine in the array.
array without modifying the turbine blades or geometry is limited. From the table it is clear that the staggered array with counter

1352
R.V. Bethi, S. Mitra and P. Kumar Renewable Energy 179 (2021) 1345e1359

Fig. 14. CT of turbines placed at a) 0.3 m b) 0.5m c) 0.7m.

Fig. 15. u of turbines placed at a) 0.3 m b) 0.5m c) 0.7m.

Table 5 current arrangement rotated in clockwise direction and generated

Power generated by 2nd turbine when placed behind 1st turbine at 0.7m 231.85W. Hence there is a change in the trend of power generation
Placement of 1st Placement of 2nd Power generated Power loss across the turbines in a staggered array. Though the power gener-
Turbine w.r.t Train Turbine w.r.t Train ated by a turbine when made to rotate in clockwise direction is
0.70 m 0.3 m 217.31 W 46.89% lesser, the array with such turbines is able to produce more power.
0.5 m 161.48 W 60.53%
0.7 m 117.02 W 71.40%
3.4.3. Case study III: 0.5m in-line array
Next, we consider an array where all the turbines are placed in-
rotating turbines is able to generate more power when compared to line at a distance of 0.5m from the train as shown in Fig. 23. A
the staggered array with turbines that rotates in the same direc- similar arrangement produced the most power compared to other
tion; see Tables 6 and 7 The total power generated by the staggered in-line arrangements when only two turbines were considered.
arrangement with counter rotating turbines is 1005.16W, which is similar to above arrangements, the present array arrangement also
158.45W more than the staggered arrangement of turbines dis- contains four turbines rotating in the same direction. Fig. 24 shows
cussed in the previous section. It is observed that the 3rd turbine the variation of CT with respect to time for each of the turbines in
produces more power than the 4th turbine, contrary to the previ- the array. The torque values between the 2nd, 3rd and 4th turbines
ous discussions where the 3rd turbine placed close to the train had are similar. As expected, there is a considerable performance dif-
to generate power in the presence of vortices emanating from both ference between the 1st turbine and the rest. But there seems to be
the train and preceding turbines. This maybe due to the fact that very little difference in performance among the 2nd, 3rd and 4th
the turbine that rotates in the counter-clockwise direction is able to turbines; see Fig. 25.
generate more power than the turbine that rotates in the clockwise Unlike the array where the turbines were arranged in a stag-
direction. This is also seen in the case of 2nd turbine. The 2nd gered fashion, there seems to be no significant drop between the
turbine in the previous arrangement rotated in counter-clockwise 2nd and 3rd turbine. Moreover, apart from the 1st turbine, the
direction and generated 248.93W, whereas the 2nd turbine in the performance of the remaining turbines are very similar and it
1353
R.V. Bethi, S. Mitra and P. Kumar Renewable Energy 179 (2021) 1345e1359

Fig. 16. Surface representation of power generation.

Fig. 17. Schematic representation of staggered array arrangement with co-rotating turbines.

Fig. 18. CT of (a) 1st turbine, (b) 2nd turbine, (c) 3rd turbine and (d) 4th turbine in the array.

seems that it has reached convergence. This would essential imply in power generation of a turbine in an array and hence such pre-
that if the array consisted of ten turbines then the performance dictions may not hold true.
drop of the 10th turbine would be approximately 60%e65%. How- Table 8 details the power generated by each turbine in the array.
ever, there are several other factors which plays an important role The total power generated by the array is 979.71W.

1354
R.V. Bethi, S. Mitra and P. Kumar Renewable Energy 179 (2021) 1345e1359

Fig. 19. u of (a) 1st turbine, (b) 2nd turbine, (c) 3rd turbine and (d) 4th turbine in the array.

Table 6 influences of vortices from two turbines. Table 9 details the power
Power generated by turbines placed in a staggered array. generated by each turbine in the array. The 2nd and 4th turbines in
Turbine in Placement of Power generated Power loss the current arrangement generates lesser power than the 2nd and
the array Turbine w.r.t Train 4th turbines from the previous arrangement; see Tables 8 and 9
1st 0.3 m 379.76 W e However, the power generated by the 3rd turbine in the current
2nd 0.7 m 248.93 W 34.45% arrangement generates more power than the 3rd turbine from the
3rd 0.3 m 97.61 W 74.30% previous arrangement. This leads us to infer that by placing a
4th 0.7 m 120.41 W 68.29%
clockwise rotating turbine ahead of a counter-clockwise rotating
turbine we can improve the performance of the counter-clockwise
rotating turbine. The total power generated by the current
3.4.4. Case study IV: 0.5m in-line array with counter rotating arrangement is 976.93W which is slightly less than the previous
turbines arrangement. From the data, one would prefer the previous
Next, we consider an array arrangement with in-line, counter arrangement of placing the turbines in-line and rotating in the
rotating turbines. The 2nd and 4th turbines are made to rotate in same direction since the difference between the turbines is less.
clockwise direction as shown in Fig. 26. The variation of CT with The current arrangement with counter rotating turbines in-line,
respect to time is shown in Fig. 27. Here too, the peaks in CT value is may produce lesser power when more than four turbines are
either positive or negative based on the rotation of the turbine. employed in the array. Such an analysis is beyond the scope of the
Hence, the 2nd and 4th turbines that rotate in clockwise direction current study.
have negative peaks while the 1st and the 3rd turbines that rotate The staggered arrangement of four turbines in an array gener-
in counter-clockwise direction have positive peaks. The perfor- ated the least amount of power out of all the arrangements dis-
mance of 2nd turbine in the current arrangement, see Fig. 28 is less cussed above. But by placing counter rotating turbines in the
with respect to the 2nd turbine in the previous arrangement. This staggered array, we were able to improve the performance such
corroborates our earlier findings that counter-clockwise rotating that it generated the most power out of all the arrangements.
turbines performs better than their clockwise rotating However, there was considerable drop in performance among the
counterparts. turbines. The farther the turbine was in the array, the lower was its
Also, it is interesting to note that the 3rd turbine performs better performance as we move towards the downstream. But in the case
than the 2nd turbine though it has to generate power under the of in-line turbine array that was placed at 0.5m from the train, the

Fig. 20. Schematic representation of staggered array arrangement with counter rotating turbines.

1355
R.V. Bethi, S. Mitra and P. Kumar Renewable Energy 179 (2021) 1345e1359

Fig. 21. CT of (a) 1st turbine, (b) 2nd turbine, (c) 3rd turbine and (d) 4th turbine in the array.

Fig. 22. u of (a) 1st turbine, (b) 2nd turbine, (c) 3rd turbine and (d) 4th turbine in the array.

Table 7 potential to harvest energy which can get amplified if we use

Power generated by turbines placed in a counter rotating staggered array.
multiple turbines in a strategic arrangement. We were able to come
Turbine in Placement of Power generated Power loss up with modifications to traditional Savonius turbine so that it is
the array Turbine w.r.t Train able to generate power from the winds produced by trains travel-
1st 0.3 m 379.76 W e ling inside tunnels. The placement of multiple such turbines in an
2nd 0.7 m 231.85 W 38.95% array to increase total power generated was investigated in detail
3rd 0.3 m 203.98 W 46.29% and an optimum arrangement was found.
4th 0.7 m 189.57 W 50.08%

C Initially, the optimum distance between two turbines in an

performance difference between the last three turbines was mini- array was determined. Farther the turbines are kept, better
mal. It should also be noted that after introducing counter rotating was their performance. But after a distance of 10 m, the
turbines, the in-line arrangement did not show any improvements. difference in performance was negligible. Hence a distance of
Thus to maximize power generation from trains in SMRT we have 10 m was chosen for further simulations.
to employ either a staggered arrangement of counter rotating tur- C In general, staggered arrangement consistently produced
bines where the turbines are placed at 0.3m and 0.7m alternatingly more power than in-line arrangement. The interaction of the
or an in-line arrangement of turbines placed at a distance of 0.5m wake vortices from the 1st turbine with the train and tunnels
from the train. walls was detrimental in the total power produced.
C The direction of rotation of the turbines were reversed
4. Conclusion alternatingly and the staggered arrangement of turbine with
counter rotating turbines were able to generate more power.
A very interesting study on VAWT arrays harvesting energy from C Interestingly, introducing counter rotating turbines into the
moving vehicles has been carried out. This paper has shown much in-line arrangement did not improve the total power

1356
R.V. Bethi, S. Mitra and P. Kumar Renewable Energy 179 (2021) 1345e1359

Fig. 23. Schematic representation of in-line array arrangement with co-rotating turbines.

Fig. 24. CT of (a) 1st turbine, (b) 2nd turbine, (c) 3rd turbine and (d) 4th turbine in the array.

Fig. 25. u of (a) 1st turbine, (b) 2nd turbine, (c) 3rd turbine and (d) 4th turbine in the array.

produced. This may be due to the reduced positive coupling advisable to employ a staggered arrangement of counter-rotating
effect between vortices and turbines. turbines placed alternatively with appropriate spacing. This study
also opens up the possibility of tremendous power generation at
The current study concludes with the valuable results on stra- any remote place as train tracks are everywhere. This green power
tegic turbine arrangements harvesting power from the train in a produced and can be used to lighten up guardhouse, transmitted to
constricted space. In order to maximize the power generation, it is nearby villages or remote areas, and can light LEDs that consume

1357
R.V. Bethi, S. Mitra and P. Kumar Renewable Energy 179 (2021) 1345e1359

Table 8
Power generated by turbines placed in an in-line array.

Turbine in Placement of Power generated Power loss

the array Turbine w.r.t Train

1st 0.50 m 428.24 W e

2nd 190.99 W 55.40%
3rd 181.67 W 57.58%
4th 178.81 W 58.25%

Fig. 26. Schematic representation of in-line array arrangement with counter rotating turbines.

Fig. 27. CT of (a) 1st turbine, (b) 2nd turbine, (c) 3rd turbine and (d) 4th turbine in the array.

Fig. 28. u of (a) 1st turbine, (b) 2nd turbine, (c) 3rd turbine and (d) 4th turbine in the array.

1358
R.V. Bethi, S. Mitra and P. Kumar Renewable Energy 179 (2021) 1345e1359

Table 9 [3] R.J. Barthelmie, K. Hansen, S.T. Frandsen, O. Rathmann, J. Schepers, W. Schlez,
Power generated by counter rotating turbines placed in an in-line array. J. Phillips, K. Rados, A. Zervos, E. Politis, et al., Modelling and measuring flow
and wind turbine wakes in large wind farms offshore, Wind Energy 12 (5)
Turbine in Placement of Power generated Power loss (2009) 431e444.
the array Turbine w.r.t Train [4] N.J. Choi, S.H. Nam, J.H. Jeong, K.C. Kim, Numerical study on the horizontal axis
turbines arrangement in a wind farm: effect of separation distance on the
1st 0.3 m 428.24 W e
turbine aerodynamic power output, J. Wind Eng. Ind. Aerod. 117 (2013)
2nd 0.7 m 182.27 W 57.44% 11e17.
3rd 0.3 m 187.67 W 56.18% [5] S. Chowdhury, J. Zhang, A. Messac, L. Castillo, Unrestricted wind farm layout
4th 0.7 m 178.75 W 58.26% optimization (uwflo): investigating key factors influencing the maximum
power generation, Renew. Energy 38 (1) (2012) 16e30.
[6] R. Ahmadian, R. Falconer, B. Bockelmann-Evans, Far-field modelling of the
hydro-environmental impact of tidal stream turbines, Renew. Energy 38 (1)
very little power. Through this study, the authors hope to shed light (2012) 107e116.
on the possibility of generating energy from unconventional sour- [7] R.J. Barthelmie, O. Rathmann, S.T. Frandsen, K. Hansen, E. Politis,
J. Prospathopoulos, K. Rados, D. Cabezo n, W. Schlez, J. Phillips, et al., Modelling
ces and highlight their potential to replace non-renewable energy and measurements of wakes in large wind farms, in: Journal of Physics:
resources. Conference Series, vol. 75, IOP Publishing, 2007, 012049, 1.
[8] M.J. Churchfield, Y. Li, P.J. Moriarty, A large-eddy simulation study of wake
propagation and power production in an array of tidal-current turbines, Phil.
CRediT authorship contribution statement Trans. R. Soc. A 371 (1985) 20120421, 2013.
[9] H. Jasak, A. Jemcov, Z. Tukovic, et al., Openfoam: a cþþ library for complex
Rajagopal Vinod Bethi: Methodology, Software, Investigation, physics simulations, in: International Workshop on Coupled Methods in Nu-
merical Dynamics, vol. 1000, IUC Dubrovnik, Croatia, 2007, pp. 1e20.
Validation, Formal analysis, Writing e original draft, preparation.
[10] H. Jasak, Dynamic mesh handling in openfoam, in: 47th AIAA Aerospace
Santanu Mitra: Conceptualization, Supervision, Resources, Writing Sciences Meeting Including the New Horizons Forum and Aerospace Exposi-
e review & editing. Pankaj Kumar: Introduction with the tools, tion, 2009, p. 341.
Guidance, Computer Implementation, Writing e review & editing. [11] J. Foo, M. Ng, L. Wang, Alstom metropolis c830, URL, http://www.sgtrains.
com/train-c830.html, Dec 2017.
[12] J. Foo, M. Ng, L. Wang, Circle line, URL, http://www.sgtrains.com/network-ccl.
Declaration of competing interest html, Oct 2017.
[13] M. Beaudoin, H. Jasak, Development of a generalized grid interface for
turbomachinery simulations with openfoam, in: Open Source CFD Interna-
The authors declare that they have no known competing tional Conference, vol. 2, 2008. Berlin.
financial interests or personal relationships that could have [14] R.E. Sheldahl, L. Feltz, B.F. Blackwell, Wind tunnel performance data for two-
appeared to influence the work reported in this paper. and three-bucket savonius rotors, J. Energy 2 (3) (1978) 160e164.
[15] D. Afungchui, B. Kamoun, A. Helali, Vortical structures in the wake of the
savonius wind turbine by the discrete vortex method, Renew. Energy 69
References (2014) 174e179.
[16] U. Saha, S. Thotla, D. Maity, Optimum design configuration of savonius rotor
[1] E.E. Commission, Shedding Light on Energy in the EU - A Guided Tour of through wind tunnel experiments, J. Wind Eng. Ind. Aerod. 96 (8e9) (2008)
Energy Statistics - 2018 Edition, Eurostat Publications - Statistical Book, 2018. 1359e1375.
[2] R.V. Bethi, P. Laws, P. Kumar, S. Mitra, Modified savonius wind turbine for [17] M. Mohamed, G. Janiga, E. Pap, D. The venin, Optimal blade shape of a modified
harvesting wind energy from trains moving in tunnels, Renew. Energy 135 savonius turbine using an obstacle shielding the returning blade, Energy
(2019) 1056e1063. Convers. Manag. 52 (1) (2011) 236e242.

1359