You are on page 1of 7

This article appeared in a journal published by Elsevier.

The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elseviers archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
chemical engineering research and design 9 1 ( 2 0 1 3 ) 221226
Contents lists available at SciVerse ScienceDirect
Chemical Engineering Research and Design
j our nal homepage: www. el sevi er . com/ l ocat e/ cher d
Comparing the energy efciency of different high rate algal
raceway pond designs using computational uid dynamics
Kurt Liffman

, David A. Paterson, Petar Liovic, Pratish Bandopadhayay


CSIRO Materials Science and Engineering, Graham Rd, Highett, Vic 3190, Australia
a b s t r a c t
High Rate Algal Ponds with long, shallow, looped channels and powered by paddlewheels have been used since the
late 1970s for growing algae to produce nutraceuticals and to remediate wastewater. These high rate or raceway
ponds are also being applied to other elds such as CO
2
capture and biodiesel production, where there is an ongoing
effort to nd ways of minimizing operating and capital costs. One approach for minimizing costs is to improve the
energy efciency of raceway ponds as this would lower operating costs and allow the construction of larger ponds,
which should also lower capital costs. A major component of energy loss in a raceway pond is the energy required
to circulate the uid around the raceway, particularly at the hairpin bends.
In order to design a low-energy consumption bend for a raceway pond, a computational uid dynamics (CFD)
code was used to model different raceway bend congurations. These simulations assisted in the development of
a novel bend design that can reduce the energy loss at a bend by around 87% relative to the energy consumed at a
conventional bend. This new design can also remove or reduce the incidence of stagnation regions within the ow.
Such a development should, potentially, improve the productivity of algae in raceway ponds.
Crown Copyright 2012 Published by Elsevier B.V. on behalf of The Institution of Chemical Engineers. All rights
reserved.
Keywords: Algae; Raceway pond design; Computational uid dynamics
1. Introduction
Most algae raceways or High Rate Algal Ponds (HRAPs) use
channels with constant width and depth, hairpin bends, and
one or more paddle wheels to circulate water through the
system. In such systems, resistance to water circulation has
two distinct parts: in the straight sections and in the hair-
pin bends (Green et al., 1995; Borowitzka, 2005). Most of the
energy loss occurs at the hairpin bends (ibid.) and a number
of modications have been developed to minimize this energy
loss.
For example, multi-vane ow rectiers have been placed
at the bends to minimize separation regions and energy loss
(e.g., Shimamatsu and Tominaga, 1980; Shimamatsu, 1987).
Another approach is to build an eccentric curved wall with
bafes (Dodd, 1986). To evaluate the effectiveness of these
and other designs requires physical modelling and/or compu-
tational uid dynamic (CFD) modelling. All of the published
analyses of energy loss in raceway ponds have been based on

Corresponding author. Tel.: +61 3 9252 6167; fax: +61 3 9252 6240.
E-mail address: Kurt.Liffman@csiro.au (K. Liffman).
Received20 September 2011; Receivedinrevisedform6 July 2012; Accepted6 August 2012
standard engineering head loss terms based on the Bernoulli
equation (e.g., Green et al., 1995). CFD analysis of raceway
ponds is only starting to appear in the literature, e.g., James
and Boriah (2010) who have developed a comprehensive, pre-
dictive model of algal growthrates coupledwitha CFDcode. To
date, there has been no published CFD analysis of the relative
efciency of different raceway pond designs. As such, there
is an opportunity for evaluating the current designs and pro-
ducing new designs which may minimize the energy loss in
raceway ponds and thereby create more efcient algal growth
systems.
Two engineering methodologies are used in this paper to
obtain the goals of removing ow separation regions and to
reduce the kinetic energy losses at sharp turns. The rst strat-
egy is to direct the water towards the outer edge of the turn
in order to keep the radius of curvature large. The other strat-
egy is to keep the channel cross-sectional area perpendicular
to the ow direction constant in order to keep the ow speed
uniform.
0263-8762/$ see front matter. Crown Copyright 2012 Published by Elsevier B.V. on behalf of The Institution of Chemical Engineers. All rights reserved.
http://dx.doi.org/10.1016/j.cherd.2012.08.007
Author's personal copy
222 chemical engineering research and design 9 1 ( 2 0 1 3 ) 221226
Fig. 1 Raceway bend geometry as shown for a standard
raceway pond. The AA line is used to view the cross section
of the pond as shown in Fig. 2.
2. Simulation
In an effort to reduce the uid ow resistance at the hair-
pin bends of a raceway, numerical modelling was carried out
for shallow turbulent ow in a rectangular channel using the
computational uid dynamics program ANSYS CFX 12. Fig. 1
denes the coordinate systemused for a typical raceway bend,
where L is the length of the transition section, R is the chan-
nel width, and r is the radius of curvature. The central divider
is assumed to be 0.2m wide. In addition, d is the local depth,
and d
0
is the standard channel depth. In this study, we have
set R=5mand d
0
= 0.3m.
2.1. Bend geometries
Sixalternative bendgeometries were examinedandcompared
with the traditional constant-width/constant-depth bend. We
also examined a bend tted with three turning vanes, and an
asymmetrical island bend similar to that reported by Sukenik
(1999). Apictorial summary of the various channel geometries
is given in Fig. 2, where we showthe cross section of the bend
at the start of the turn, along the AA line shown in Fig. 1
where the observer is facing the straight section of the race-
way. Aquantitative descriptionof the bendgeometries is given
in Table 1.
Two new approaches to designing the bends were con-
sidered: the rst was to keep the channel width the same,
but direct the ow to the outer side of the bend by varying
the depth from shallow in the centre to deep at the outside
Fig. 2 Channel cross-sections that were examined in this
paper (not to scale). The water depth is exaggerated to
show the proles more clearly. All cross-sections have the
same area, and are taken just at the start of the turn, where
the observer is facing towards the straight section of the
raceway. (a) Standard raceway pond cross section with
m=0 (see Eq. (1)); (b) square-root prole with m=0.5; (c)
linear prole with m=1; (d) quadratic prole with m=2; (e)
wide box; (f) mediumbox; (g) narrow box.
Fig. 3 Channel width at the bend as given by the cosine
function.
(Fig. 2(b)(d)). The depth prole was assumed to be given by
the relationship:
d
d
0
= (m+1)

r
R

m
(1)
where m is a free parameter.
The second approach was to narrowthe channel width and
increase the depth, but keep the depth uniform across the
width (Fig. 2(e)(g)). The channel width in the transition sec-
tion was varied using a cosine function (Fig. 3), with the local
channel depth calculated fromthe width.
In both cases, a transition section keeping the cross-
sectional area close to constant was used from the straight
Table 1 Summary of channel bend geometries with the geometry of the straight section also included (R=5m,
d
0
=0.3m).
Channel description L/R d/d
0
a Normal design 1 1
b Square-root prole 1 1.5(r,R)
0.5
c Linear prole 1 2r,R
d Quadratic prole 2 3(r,R)
2
e Wide box 2

0 for r,R - 0.4


5,3 for r,R 0.4
f Mediumbox 2

0 for r,R - 0.5


2 for r,R 0.5
g Narrow box 2

0 for r,R - 0.6


2.5 for r,R 0.6
h Turning vanes 0 1
i Asymmetric island 2 1
Straight section (96m) 1
Author's personal copy
chemical engineering research and design 9 1 ( 2 0 1 3 ) 221226 223
standard channel to the modied section at the start of the
bend. The depth prole was retained unaltered through the
turn, and then a similar transition section was used to return
to the straight standard channel.
In addition to these approaches, two other arrangements
were included for comparison (see Table 1): arrangement h
has the standard section prole with three turning vanes
included; arrangement i has a constant depth, but incorpo-
rates an asymmetric island in each hairpin bend, similar to
that described by Sukenik (1999) and as installed at the Nature
Beta Technologies Ltd. plant in Eilat, Israel.
2.2. Modelling assumptions
The paddlewheel was represented by a momentumsource in
a rectangular volume encompassing the full depth and width
of the channel and extending a distance of 0.5m along the
channel. A source strength of 1kgm
2
s
2
was found to give
ow velocities that were close to 0.03ms
1
. Stronger sources
were used to conrmthe scaling lawof source strength versus
ow rate up to 0.3ms
1
.
All the computations inwhichthe depthof the channel was
varied were performed on an unstructured tetrahedral mesh.
For all the computations inwhichthe depthof the channel was
constant, an unstructured hexahedral mesh was used with
an aspect ratio that allowed for ner spacing in the vertical
direction than in the horizontal direction. Typically, 1,000,000
elements were used for the tetrahedral meshes and 300,000
elements were used for the hexahedral meshes.
Grid independence was checked by reanalysing the best
case (the mediumbox section) with a ner grid of 2.6 million
elements. Results showed only a 7% change in the computed
value of the channel ow rate, Q. At least four elements were
used in the vertical direction.
2.3. Ground and surface boundary conditions
The ground boundary conditions used, i.e., scalable wall func-
tion boundary conditions, are those recommended by ANSYS
CFX. There was no attempt to model the ow right up to the
wall, or to model the sublayer adjacent to the wall. Instead,
the calculation started at the boundary between the sublayer
and the turbulent boundary layer at a y
+
of 11.06, where y
+
is the non-dimensionalized distance fromthe wall. The scal-
able wall function boundary condition calculates the friction
velocity, u
t
fromthe equation:
U
t
=

u
t
k

ln(y

) +5.2 LB (2)
U
t
is the known velocity tangent to the wall at distance Ly
fromthe wall, where Ly is the near-wall grid spacing, k = 0.41
is the Von Karman constant,
y

= max

11.06.
u

Ly
v

(3)
v is the kinematic viscosity, and
u

0.3k (4)
where k is the known turbulent kinetic energy at distance Ly
fromthe wall. For smooth wall owLB=0. For rough wall ow
with sand grain roughness h:
LB =

1
k

ln

1 +0.3nu
t
v

(5)
This is usedto calculate the absolute value of the wall shear
stress t

from:
t

= ,u

u
t
(6)
where , is the uid density. The turbulent energy dissipation
is calculated from:
=
u

(0.3k)
1.5
kvy

(7)
Unlike previous wall functions that gave results that were
grid dependent (Launder, 1984), the present wall functions
have the advantage that any changes in grid resolution in
the near-wall region have no inuence on the accuracy of the
solution.
Symmetry boundary conditions were used for the top (sur-
face) boundary, because the shallow depth and large plan
area made use of multiphase airwater modelling unfeasi-
ble. Symmetry boundary conditions are accurate enough for
the present comparisons, although they do force an articial
constraint on the continuity equation.
Computations were done for both a smooth bed and for a
rough bed with a sand grain roughness of 1mm. The smooth
bed may be appropriate for a plastic-lined channel and the
rough bed for a concrete-lined channel.
2.4. Turbulence, parameters and convergence
The k turbulence model is reported to be notoriously unre-
liable in calculating ows around hairpin bends (e.g., Zijlema
et al., 1995) in deep channels. For shallow channels the inu-
ence of the bottom of the channel dominates the turbulence
generation and dissipation, and so the k turbulence model
can be used.
The parameter ratio L/R (where L is the length of the transi-
tion zone and R is the radius of the bend) was varied between
values of 1 and 2. The depth prole was given by a power law.
Changing the value of L/R from1 to 2 for the quadratic depth
prole resulted in only a slight increase (5%) in the predicted
ow rate.
Most computations converged completely to a root mean
square error in all variables of 10
6
or better. An exception
was the quadratic bed prole, where the error reached an
oscillating state, possibly indicative of slight ow instability.
Convergence was faster for the rough bed calculation than the
smooth bed calculation.
3. Simulation results
3.1. Secondary ow overturning
For most algae in a raceway pond, turbulent diffusion is
needed to continually bring the algae, that would otherwise
be shaded in the bottom of the channel, to the top of the
channel, where they can get enough light to photosynthe-
size. Secondary ows resulting in overturning also help to
bring algae to the surface fromthe lower levels. Overturning is
Author's personal copy
224 chemical engineering research and design 9 1 ( 2 0 1 3 ) 221226
generated in corners as the faster upper water is directed
towards the outer edge of the channel, and slower moving
water fromnear the bottom, up-wells near the centre.
If 180

is taken to be complete overturning from top to


bottom, and 360

for top to bottom and then back to top, all


the box sections achieved 130

of overturning for each hair-


pin bend, the vaned design achieved 200

of overturning,
the asymmetric island design achieved 45

of overturning
for those algae that did not get entrained in the recirculation
region, and the standard constant depth design achieved only
10

of overturning for those algae that did not get entrained


in the recirculation region. All estimates of overturning were
obtained by streamline tracking and are approximate esti-
mates only. To obtain a more precise estimate of overturning
would require a separate study where the CFD simulations
were run with more grid elements in the vertical direction.
3.2. Pressure loss
Fig. 4 shows the calculated normalized pressure losses for
clockwise ow in all geometries, in which red is high pres-
sure and blue is low pressure. The abrupt change in colour at
the upper left in each image is the location of the momentum
source which simulated the force provided by a paddlewheel.
The increase in pressure with distance along the straights in
Fig. 4(a) illustrates the conversion of kinetic energy into pres-
sure energy. Contours are set to divide the pressure range into
50 equal intervals.
3.3. Fluid speed variation
Normalized carrier uid ow speeds for clockwise ow in all
geometries are shown in Fig. 5. As good performance is indi-
cated by the absence of localized high uid velocity (shown
in red), the absence of large regions of low velocity (blue)
and the absence of recirculation regions (bounded by a blue
minimumbetween two other colours), the geometries shown
in Fig. 5(a)(c) and (i) performed poorly.
3.4. Hydraulic power for different bend geometries
Fig. 6 and Table 2 show the relative contributions to over-
all channel hydraulic power consumption of the ow in the
straight section of a raceway and that in the two bends for the
various bend designs. The calculations for the straight sec-
tion are based on a 96mlong raceway with a cross-section of
4.9m0.3m(1.47m
2
), and for a mean velocity of 0.3ms
1
.
It can be seen fromFig. 6 and Table 2 that the mediumbox
and narrow box bend designs produce the lowest power con-
sumption. The power consumptioninthe mediumandnarrow
box bend designs is approximately 13%of the normal or stan-
dard bend design. If one considers the bends and the straight
sections of the pond then the power loss for the mediumand
narrow box design is approximately 32% of the normal pond
design.
To compute these power consumption data, we have
applied a 100Pa pressure drop around each raceway circuit.
This pressure drop is obtained by simulating a paddle wheel
with a strength of 200Pam
1
, where the paddle wheel has
a width of 0.5m. Each particular bend conguration is tested
in the computation raceway and the ow rates computed. To
compare the power consumptionfor all the raceways, we scale
the pressure drop to a ow speed of 0.3ms
1
by using the
formula
LP
sculc
=

0.3ms
1
Q
comp
,A

2
100Pa (8)
where Q
comp
is the computed ow rate for a particu-
lar bend conguration and A is the cross sectional area
(1.47m
2
5m0.3m) for the straight section of the raceway
Fig. 4 Normalized pressure losses for clockwise ow in all geometries, in which red is high pressure and blue is low
pressure. A scale bar is provided in relative units with 1 representing the maximumpressure value and 0 representing
the minimumpressure value. (For interpretation of the references to colour in text, the reader is referred to the web version
of the article.)
Author's personal copy
chemical engineering research and design 9 1 ( 2 0 1 3 ) 221226 225
Fig. 5 Normalized ow speeds at the surface of clockwise ow in all geometries, in which red is high velocity and blue is
low velocity. A scale bar is provided in relative units with 1 representing the maximumspeed value and 0 representing
the minimumspeed value. (For interpretation of the references to colour in text, the reader is referred to the web version of
the article.)
Fig. 6 Channel hydraulic power consumption for a rough
96mlong straight section and for the various channel
hairpin bend geometries. The color graph is available from
the web.
pond. The power consumption, W
cons
, is then computed via
the formula
W
cons
= ALP
sculc
0.3ms
1
(9)
The actual energy saving for a particular raceway will be
dependent on the length of the straight sections of the race-
way. A longer pond implies that more energy is consumed
moving the water along the straight sections and the relative
energy saving (fromthe more efcient bends) decreases as the
pond length increases.
4. Economics and algal productivity
The results in the previous section demonstrate that by adopt-
ing the narrow, medium or wide box designs, it is
possible to reduce the energy consumption in the given race-
way pond by around 60%. This is a considerable hydrodynamic
energy saving, but in terms of the actual electricity-related
operating expenses this does not amount to a signicant
reduction in operating costs. This is due to the cost of electric-
ity for the paddlewheel(s) which contributes approximately
1% of total operating expenses (Benemann et al., 1987). Thus,
if we concentrate only on electricity consumption, the intro-
ductionof the new, energy efcient, raceway bends will reduce
total operating expenses by approximately 0.5%.
Focussing on electricity consumption, however, neglects
other, less obvious nancial savings. These arise, because the
ponds can now be at least twice as long for the same num-
ber of paddle wheels, so there are potential savings in capital
costs, electrical engineering and other economies of scale. As
an illustration of the latter effect, Benemann et al. (1987) sug-
gest that a factor of 10 increase in the area of a raceway pond
facility decreases the per kg (dry weight) cost of algae by 75%.
This may be a somewhat optimistic estimate, but if the scal-
ing is approximately linear then it follows that increasing the
size of the raceway ponds by a factor of two implies a poten-
tial reduction in the per kg cost the algae of order 5%. Of
course, building a more complex pond design will probably
add to the capital cost of construction. We are hoping to obtain
quantitative estimates for this extra cost from commercial
operations that hope to build ponds with the designs given in
this paper.
Finally, as is evident from Fig. 5(e)(g), the new box
designs completely remove the low speed and stagnation
regions that are present at the ends and along the cen-
tral divider of the standard raceway (Fig. 5(a)). As such,
these energy efcient designs should also improve mixing
and have the potential for improving the productivity of the
algae.
Author's personal copy
226 chemical engineering research and design 9 1 ( 2 0 1 3 ) 221226
Table 2 Channel bend geometries with corresponding ow rate and hydraulic power. Q values are ow rates that give a
100Pa pressure drop through the whole simulated raceway. Power values are for power dissipated in the two bends
alone, where the total power loss in the two straight sections is given separately. The power loss has been scaled to the
situation where the ow speed is 0.3ms
1
(i.e., when Q=0.441m
3
s
1
).
Channel
description
Smooth bed ow
rate Q (m
3
s
1
)
Rough bed ow
rate Q (m
3
s
1
)
Rough bed hydraulic
power (W)
a Normal design 0.248 0.273 86
b Square-root prole 0.317 0.276 83
c Linear prole 0.358 0.304 64
d Quadratic prole 0.418 0.359 38
e Wide box 0.562 0.450 13
f Mediumbox 0.577 0.463 11
g Narrow box 0.578 0.467 11
h Turning vanes 0.356 0.315 58
i Asymmetric island 0.342 0.290 73
Straight 96m 0.606 0.498 35
5. Conclusions
To minimize energy consumption in raceway ponds, we have
proposed a number of 3-D bend geometries that deect
ow to the outer edge of the bend and retain the channel
cross-sectional area, thereby minimizing energy losses due
to centrifugal forces. Numerical modelling using ANSYS CFX-
12 was used to model the new bend congurations and also
three existing bend designs. The results demonstrate that the
novel bend designs can cut energy losses by 87% compared to
a conventional bend, and by 82%compared to a vaned design.
Some of the new bend designs, in particular the narrow,
medium and wide box designs not only minimize energy
consumption, but also improve the mixing of the raceway
pond by removing low speed and stagnation regions within
the ow. Such a development should, potentially, improve the
productivity of algae in raceway ponds.
Finally, the higher energy efciency of the new raceway
pond designs will allow the construction of larger ponds with
lower relative capital costs thereby potentially producing an
approximate 5% saving in the nal production cost of algae.
Acknowledgements
We wish to acknowledge funding from the CSIRO Energy
Transformed Flagship and the constructive reviews from the
anonymous referees.
References
Benemann, J.R., Tillett, D.M., Weissman, J.C., 1987. Microalgae
biotechnology. Trends Biotechnol. 5, 4753.
Borowitzka, M.A., 2005. Culturing microalgae in outdoor ponds.
In: Andersen, R.A. (Ed.), Algal Culturing Techniques. Elsevier
Academic Press, Burlington, pp. 205218.
Dodd, J.C., 1986. Elements of pond design and construction. In:
Richmond, A. (Ed.), CRC Handbook of Microalgal Mass Culture.
CRC Press, Boca Raton, pp. 265283.
James, S.C., Boriah, V., 2010. Modeling algae growth in an
open-channel raceway. J. Comput. Biol. 17, 895906.
Green, F.B., Lundquist, T.J., Oswald, W.J., 1995. Energetics of
advanced integrated waste-water pond systems. Water Sci.
Technol. 31, 920.
Launder, B.E., 1984. Numerical computation of convective heat
transfer in complex turbulent ows: time to abandon wall
functions? Int. J. Heat Mass Transfer. 27, 14851491.
Shimamatsu, H., 1987. A pond for edible spirulina production and
its hydraulic studies. Hydrobiologia 151, 8389.
Shimamatsu, H., Tominaga, Y., 1980. Apparatus for cultivating
algae. U.S. Patent Ofce, Patent No. 4217728.
Sukenik, A., 1999. Production of eicosapentaenoic acid by the
marine eustigmatophyte nannochloropsis. In: Cohen, Z. (Ed.),
Chemicals fromMicroalgae. Taylor & Francis, London, pp.
4156.
Zijlema, M., Segal, A., Wesseling, P., 1995. Finite-volume
computation of incompressible turbulent ows in general
coordinates on staggered grids. Int. J. Numer. Method Fluid 20,
621640.

You might also like