Desalination 433 (2018) 41–47

Contents lists available at ScienceDirect

Desalination
journal homepage: www.elsevier.com/locate/desal

Numerical analysis of performance of ideal counter-current flow pressure T

retarded osmosis
⁎
Wenjuan Yanga,b, Lianfa Songc, , Jianqiang Zhaoa,b, Ying Chena,b, Bo Hud
a
Key Laboratory of Subsurface Hydrology and Ecological Effects in Arid Region, Ministry of Education, Chang'an University, Xi'an, Shaanxi 710054, P.R. China
b
School of Environmental Science and Engineering, Chang'an University, Xi'an, Shaanxi 710054, P.R. China
c
Department of Civil, Environmental, and Construction Engineering, Texas Tech University, Lubbock, TX 79409-1023, United States
d
School of Civil Engineering, Chang'an University, Xi'an, Shaanxi 710061, PR China

G RA P H I C A L AB S T R A C T

A R T I C L E I N F O A B S T R A C T

Keywords: Pressure retarded osmosis (PRO) operated in counter-current flow mode is more efficient than in co-current flow
Pressure retarded osmosis mode to extract salinity gradient energy. Knowledge of the performance of counter-current flow PRO under
Counter-current flow various operating conditions (on equilibrium or off equilibrium) is of paramount importance to understand the
Optimal technique potential capacity of the technology and to optimize process design. In this study, a systemic and rigorous
Numerical procedure
numerical procedure was developed for performance simulation of counter-current flow PRO. An optimization
Performance analysis
technique was used to accurately determine the originally unknown flow rate of the draw solution at the feed
entrance of membrane channel so that the procedure could also be used for PRO systems not at equilibrium. With
this numerical procedure, new interesting findings were made about the ideal counter-current flow PRO. A
characteristic parameter of the PRO, the required membrane area to reach equilibrium for any given operating
condition, was determined and reported for the first time. Another exciting finding was that the no-flux zone
(dead region) occurs adjacent the draw entrance at the critical feed fraction when the membrane area is greater
than the required equilibrium area. Power density and specific energy in PRO under various conditions were
investigated with this numerical procedure.

1. Introduction
A huge amount of energy is dissipating every day globally when
river waters from the inland mix with seawater [1]. This clean and
renewable energy is called salination power [2], osmotic power [3], or
most often salinity gradient energy [4]. The global fresh water run-off is
about 1.1 × 106 m3/s that contains salinity gradient power of
2.6 × l012 watts with reference to seawater [5]. In theory, up to 0.8 kW
per cubic meter of fresh water is extractable in controlled mixing with
seawater — equivalent to the potential energy contained in water of
280 m high hydraulic head [1,6].
Pressure retarded osmosis (PRO) [7,8] is the most studied

⁎
Corresponding author.
E-mail address: lianfa.song@ttu.edu (L. Song).

https://doi.org/10.1016/j.desal.2017.12.058
Received 5 June 2017; Received in revised form 30 October 2017; Accepted 26 December 2017
Available online 03 February 2018
0011-9164/ © 2018 Elsevier B.V. All rights reserved.
W. Yang et al. Desalination 433 (2018) 41–47

Nomenclature QFS flow rate of feed solution at the draw entrance

QLB lower bound of the draw flow rate range at the feed en-
A membrane permeability trance
cD salt concentration of draw solution QUP upper bound of the draw flow rate range at the feed en-
cD0 salt concentration of draw solution at the feed entrance trance
cDS initial salt concentration of draw solution at the draw S total membrane area
entrance s variable for membrane area
cF salt concentration of feed solution SE specific energy
cF0 initial salt concentration of feed solution at the feed en-
trance Subscripts
cFS salt concentration of feed solution at the draw entrance
cD∗ salt concentration of draw solution at equilibrium i the ith step from the feed entrance
cF∗ salt concentration of feed solution at equilibrium n total step number
fos osmotic pressure coefficient
Jw water flux Greek symbols
PD power density
QD flow rate of draw solution θ feed fraction
QD0 flow rate of draw solution at the feed entrance θ⁎ critical feed fraction
QDS initial flow rate of draw solution at the draw entrance π osmotic pressure
QF flow rate of feed solution ΔP retarded pressure
QF0 initial flow rate of feed solution at the feed entrance Δs step size along membrane channels

technology in harnessing salinity gradient energy and widely con- developed for modeling and simulation of counter-current flow PRO.
sidered more promising than the alternative reverse electrodialysis Bisection method of optimization technology was used to accurately
(RED) [9,10]. A semipermeable membrane that allows only water to determine the unknown boundary conditions on one end of the chan-
pass is employed in PRO to extract salinity gradient energy. Two so- nels that match the given values on the other end. The numerical
lutions of different salinities are separated by the semipermeable procedure can be used to simulate the performance of counter-current
membrane with a hydraulic (retarded) pressure smaller than the os- PRO under any conditions regardless equilibrium state. The new nu-
motic pressure difference added on the higher salinity (draw solution) merical procedure enables the investigations on many interesting as-
side of the membrane. Osmotic flow of water from the low salinity (feed pects of the counter-current flow PRO that could not be done pre-
solution) side expands the volume of draw solution on the high salinity viously.
side. The increased volume under the hydraulic pressure can be con-
verted to electricity with a turbine. Feed solution in PRO can be river,
brackish water, or waste water while draw solution can be seawater or 2. Mathematical model and numerical procedure
brines of high salinities [11–15].
Efficiency of PRO is an important consideration for economically 2.1. Mathematical model of counter-current flow PRO
harvest and utilization of salinity gradient energy [16,17]. There are
basically two operation modes of the module scale PRO process: A counter-current flow PRO module is schematically shown in
counter-current flow and co-current flow. It can be shown that counter- Fig. 1. A semi-permeable membrane separates the module into two
current flow mode is more efficient in extracting salinity gradient en- channels. Draw solution enters at the right end of a channel flowing to
ergy than co-current flow mode [18–20]. Knowledge of the thermo- the left while feed solution enters at the left end of the other channel
dynamic limit on the performance of counter-current flow PRO system flowing oppositely. Water transports through the membrane from the
and the affecting factors is of critical importance to assess the viability feed solution to draw solution under the net driving pressure, which is
of the technology and to optimize the process [18]. the osmotic pressure difference minus the hydraulic pressure added on
A numerical procedure that is capable to model and simulate the the draw solution side. The increase in draw solution flow rate (or
performance of the counter-current flow PRO system under any oper- decrease in feed solution flow rate) along the module is a primary
ating conditions would be a powerful tool for viability assessment of
PRO. One particular difficulty for such numerical solution is that ade-
quate boundary conditions cannot be specified at either side of the
counter-current flow PRO. To start the numerical calculation, there are
four parameters including concentrations and flow rates of both the
feed solution and the draw solution are needed in one side. However,
on one side of counter-current flow PRO, only the stream into the
membrane module is of known flow rate and salt concentration while
the flow rate and salt concentration of the stream out of the module are
determined as a result of PRO performance. To overcome this difficulty,
an assumption of sufficiently large membrane area was employed in the
previous studies so that either the draw limiting regime or the feed
limiting regime was reached. In that case, the flow rate and salt con-
centration of outgoing stream of PRO module can be determined by
simple mass balance calculation. However, the method cannot be used Fig. 1. A schematic of counter-current flow PRO. Feed and draw solutions of known flow
rates (QF0 and QDS) and salt concentrations (cF0 and cDS) are provided to feed entrance
for the general cases without knowledge of attainment of the limiting
s = 0 and draw entrance s = S, represented. The flow rate (QFS and QD0) and salt con-
regimes.
centrations (cFS and cD0) at the opposite ends of the membrane channels are unknowns.
In this study, a systemic and rigorous numerical procedure was ΔP and Jw are the retarded pressure and the water flux, respectively.

42
W. Yang et al.
parameter to calculate the energy output of the PRO module.
With the assumptions of perfect salt rejection and absence of con-
centration polarization, the governing equations for flow rate changes
along the membrane channels in an ideal counter-current flow PRO
module can be constructed by applying mass conservation principle to
draw and feed solutions [19] as:
dQD
= −JW (cD , cF , ∆P )
ds
dQF
= −JW (cD , cF , ∆P )
ds
where QD and QF are the flow rates of draw and feed solutions, re-
spectively, cD and cF are the salt concentrations of draw and feed so-
lutions, respectively, s is the variable for membrane area from the feed
entrance, Δ P is the hydraulic (retarded) pressure applied on draw so-
lution, and Jw is the water flux through the membrane, which is a
function of the salt concentrations in draw and feed solutions and the
retarded pressure.
While the retarded pressure is a constant in the PRO, salt con-
centrations change along the membrane channels as a result of water
transport from feed solution to draw solution. Because a complete salt
rejection is assumed for the membrane, salt concentrations in both
draw and feed channels can be easily determined from the mass balance
relationship of salt, as:
cDS QDS
cD =
QD
cF 0 QF 0
cF =
QF
where QDS and QF0 are the initial flow rates of draw and feed solutions,
respectively, and cDS and cF0 are the initial salt concentrations of draw
and feed solutions, respectively.
Because of the assumption of absence of internal concentration
polarization and complete mixing in the channels, water flux can be
directly related to the difference of bulk concentrations of draw and
feed solutions and the retarded pressure as:
JW = A [π (cD ) − π (cF ) − ∆P ] = A [fos (cD − cF ) − ∆P ]
where A is the water permeability of membrane, π is the osmotic
pressure, and fos is the osmotic pressure coefficient that is dependent on
solution composition and temperature. A linear relationship between
the osmotic pressure and the salt concentration as shown in Eq. (5) is
usually adequate.
In a counter-current flow PRO, because draw solution and feed so-
lution enter the membrane channels from opposite ends, the initial
boundary conditions for Eqs. (1) and (2) are specified as:
QF = QF0, at s = 0
QD = QDS , at s = S
where S is the total membrane area in the PRO channels. In this study,
the entrance of feed solution is selected as the start of the channels, i.e.,
s = 0. On the other hand, s = S corresponds to the end of the channels,
at which draw solution enters the channel flowing backward. Together
with flow rates, salt concentrations cF0 and cDS of feed and draw solu-
tions are also known at their own entrances.
As indicated in Fig. 1, flow rate and salt concentration of draw so-
lution at the feed entrance and the flow rate and salt concentration of
feed solution at the draw entrance of the membrane channels are un-
knowns. In fact, a principal purpose to solve the governing equations is
to find out these values. A particular difficulty in this problem is that
these values are also the required boundary conditions to solve the
governing equations. To overcome this difficulty, an iterative numerical
solution procedure is developed below.
Desalination 433 (2018) 41–47
2.2. Numerical procedure
In this numerical solution procedure, the membrane with a total
area of S is divided into n equal segments of step size Δ s = S/n. When n
is sufficiently large, the variables on each segment can be reasonably
approximated with constant values. The feed entrance (the draw exit)
on the left end of the channels is denoted as the starting point i = 0,
and grids are continuously numbered with i = 1, 2, 3, …, until n for the
(1)
ending point at the feed exit (the draw entrance) on the right of the
channels.
Eqs. (1) and (2) in the forward finite difference form are rewritten
(2)
as:
QD, i + 1 = QD, i − JW , i ∆s (8)
QF , i + 1 = QF , i − JW , i ∆s (9)
Numerical calculation starts from i = 1 and stops at i = n. At any
step, salt concentrations can be determined with Eqs. (3) and (4) when
flow rates become available from the numerical solutions. Then water
flux through the membrane can be readily calculated with Eq. (5).
Flow rate of draw solution at the starting point QD, 0 is required to
start the numerical calculation. Only in the special cases at equilibrium,
the value of QD, 0 can be determined by equal the osmotic pressure
difference to the retarded pressure. Draw flow rate is generally un-
available from the problem itself because it is initially specified on the
draw entrance of the membrane channels. The value of draw flow rate
at the feed entrance can only be determined after the solution of the
(3) problem has been obtained. A mathematical programing technique
(bisection method) will be used to find the right draw flow rate at the
starting point and numerical solution of the problem.
(4) A proper range is first selected for all feasible draw flow rates at the
feed entrance. The lower bound of the range is obviously the initial flow
rate of draw solution entering the module:
QLB = QDS (10)
where QLB is the lower bound of the draw flow rate range at the feed
entrance. When equilibrium is established at the feed entrance, the
upper bound of the draw flow rate range is
QDS cDS
(5) QUB =
cF 0 + ∆P / fos (11)
where QUB is the upper bound of the draw flow rate range at the feed
entrance. Eq. (11) works for the cases where there is sufficient supply of
water from feed solution. Equilibrium may not be reached at the feed
entrance due to the limited water supply from feed solution. In this
case, a natural limit on the draw flow rate at the feed entrance is the
sum of the initial draw flow rate and initial feed flow rate as:
QUB = QDS + QF0 (12)
(6) The smaller one of the values calculated with Eqs. (11) and (12) will
be chosen as the upper bound QUB of the draw flow rate range at the
(7) feed entrance. The mistakenly use of the larger one as the upper bound
can lead to unreasonable results.
Draw flow rate at the feed entrance can then be determined by
applying bisection method to the flow rate range. Bisection method is
an iteration procedure that reduces the range width by half every
iteration. Iteration stops (the right flow rate is found) when the range
width becomes smaller than the preset tolerance. In each iteration, the
middle point of current range:
QUB + QLB
QD,0 =
2 (13)
is used as the trial draw flow rate at the starting point i = 0 to carry out
the numerical solution with Eqs. (8) and (9).
The calculated draw flow rate at the ending point (the draw en-
trance) QD, n will be compared with the specified boundary condition
QDS. If the calculated draw flow rate is equal to the boundary condition
43
W. Yang et al.
within the allowable tolerance, the trial draw flow rate is accepted as
the true value at the feed entrance. If the calculated draw flow rate at
the draw entrance is not equal to the given boundary condition, the
range width will be reduced by half. When the calculated value is
greater than the boundary condition, the upper bound QUB will be set to
QD,0 while the lower bound QLB remains unchanged. Otherwise, the
lower bound QLB will be set to QD,0 while the upper bound QUB remains
unchanged. The middle value of the updated flow rate range is then
used for the next trial calculation until the right draw flow rate at the
feed entrance is found. The right draw flow rate at the feed entrance is
guaranteed to be found in a limited number of iterations.
Using the numerical procedure for Eqs. (1) and (2), draw and feed
flow rates along the membrane channels can be obtained at the same
time the right draw flow rate at the feed entrance is identified. Then the
distributions of salt concentrations of both solutions and transmem-
brane water flux along the membrane channels can be calculated with
Eqs. (3)–(5). The numerical solution procedure was coded in Visual C+
+ language and all the numerical solutions could be obtained on a
common personal computer within a second.
3. Results and discussion
In this section, the numerical solution procedure is first demon-
strated to work well for counter-current flow PRO under any conditions
without requirement for equilibrium at either end of the membrane
channels. Then more numerical simulations are carried out to study the
features of counter-current flow PRO under various conditions. The
performance of counter-current flow PRO is assessed with power den-
sity and specific energy based on the numerical solutions. All simula-
tions are carried out with the following default parameter values in
Table 1 unless other stated. The osmotic pressure coefficient for sea-
water 73 Pa·L/mg is used to calculate osmotic pressure from salt con-
centration. The membrane area 10 m2 given in Table 1 is the smallest
area used in the simulations. It will be specified when larger membrane
areas are used. In all the simulations, the feed flow rate is fixed while
the draw flow rate varies with the feed fraction.
3.1. Features of mass transfer in PRO with small and large feed fractions
Feed fraction is commonly used to indicate the relative amounts of
feed and draw solutions in a PRO process, which is defined as:
QF 0
θ=
QF 0 + QDS (14)
where θ is the feed fraction. Numerical solutions of PRO operated at
small and large feed fractions are carried out and discussed for their
distinct features in this section. These numerical solutions also serve to
demonstrate the effectiveness of the numerical procedure.
For a small feed fraction θ = 0.2, the flow rates from the numerical
solution, as well as corresponding salt concentrations and water flux,
along the membrane channels are presented in Fig. 2 for different
membrane areas, with solid lines for draw solution and dashed lines for
feed solution in Fig. 2a and b. It can be seen from Fig. 2 that equilibrium
is not reached in the membrane channels with the membrane areas of
20 m2 and 40 m2 (curve 1 and curve 2). Flow rates and salt con-
centrations of feed and draw solutions vary continuously along the
whole channels. Water flux increases monotonically throughout the
channels. On the contrary, equilibrium is reached for membrane areas
of 80 m2 and 100 m2 (curve 3 and curve 4) at about 75 m2 in the
channels. The equilibrium is indicated by flat portions in the curves in
Fig. 2a and b for constant flow rates and salt concentrations near the
draw entrance of the channels. Fig. 2c shows that water flux first in-
creases to a peak value and then drops to zero in the portion of equi-
librium for membrane areas of 80 m2 and 100 m2.
Because water flux across the membrane at equilibrium is zero at
the draw entrance, salt concentration in the feed stream cF∗ leaving the
44
Desalination 433 (2018) 41–47
module can be determined from Eq. (5):
∆P
cF∗ = cDS −
fos (15)
Numerical simulations for a PRO with a large feed fraction θ = 0.8
are also carried out and the results are shown in Fig. 3 with the solid
lines are for draw solution and dash lines for feed solution in Fig. 3a and
b. Because major changes occur near the draw entrance, the results are
presented with the draw entrance as origin for all total membrane
areas.
One striking feature in Fig. 3 is that the curves for different total
membrane areas almost coincide with each other near the draw en-
trance with major changes occurring in the first 40 m2. Flow rates, salt
concentrations all change relatively quickly in this region near the draw
entrance and then approaching gradually to certain values. As indicated
by the curves in Fig. 3c, water flux starts with the maximum value at
the draw entrance and decreases monotonically toward to the feed
entrance. In the region of significant flux, the total membrane area has
little impact on the water transfer. Another feature is that water flux
changes much more moderately toward equilibrium point (zero flux)
than in the cases for small feed fraction. The membrane area for equi-
librium to be reached is about 64 m2. Because a relatively small fraction
of water transfers from feed stream to draw stream with a large feed
fraction, the changes in the salt concentrations of the feed stream is
insignificant along the whole membrane channel. If equilibrium is
reached, salt concentration of the draw solution cD∗ leaving the mem-
brane module can be calculated with Eq. (5):
∆P
cD∗ = cF 0 +
fos (16)
A priori knowledge about the existence of equilibrium is usually
unavailable when a counter-current flow PRO is under consideration.
The numerical procedure presented here guarantees solutions for any
cases without the need for a priori knowledge or assumption. All si-
mulations can be completed within 10–30 iterations dependent on the
used tolerance criterion.
3.2. Behavior under critical feed flow rate fraction
It has been demonstrated in Figs. 2 and 3 that equilibrium occurs
near the entrance of draw solution for small feed fraction θ but near the
entrance of feed solution for large feed fraction θ if the membrane area
is sufficiently large. It is more striking that water flux behaves com-
pletely different as it approaches to the equilibrium in the two situa-
tions. For small feed fraction, water flux from a nonzero value at the
feed entrance increases to a peak near the front of equilibrium point
and then drops quickly to zero at the equilibrium point, leaving the
remaining membrane channels to the draw entrance with zero water
flux (dead region). In contrast, water flux for the large feed fraction
starts from a nonzero value at the draw entrance decreases gradually to
zero at the equilibrium point and the region of zero flux occurs adjacent
the feed entrance.
It is reported [18] that there is a critical feed fraction at which
equilibrium occurs on both ends of the module simultaneously. This
Table 1
Default parameter values in numerical simulations.
Parameter Value Unit
Draw solution salinity 35,000 mg/L
Feed solution salinity 500 mg/L
Osmotic pressure coefficient 73 Pa·L/mg
Retarded pressure 1.45 × 106 Pa
Membrane permeability 2 × 10− 11 m/s·Pa
Membrane area 10 m2
Feed flow rate 0.001 m3/s
W. Yang et al.
a
b b
c c
Fig. 2. (a) Flow rates, (b) salt concentrations, and (c) permeate flux in counter-current
flow PRO with the feed fraction θ = 0.2. The solid curves in (a) and (b) represent the
draw solution and the dashed curves represent the feed solution. Numbers 1, 2, 3, 4
indicate the total membrane area of 20 m2, 40 m2, 80 m2, and 100 m2, respectively.
critical feed fraction θ⁎ is given by:
cF∗
θ∗ =
cDS + cF0 (17)
The PRO system with the critical feed fraction at equilibrium re-
presents a case for extraction of the maximum specific energy from the
salinity gradient. The features of such a PRO system, such as the
membrane area to reach equilibrium, behavior of water flux, and lo-
cation of region of zero flux, are of particular interests and will be in-
vestigated here.
From the parameters given in Table 1, the critical feed fraction
θ⁎ = 0.426 is obtained. Using this value, the numerical solutions of the
counter-current flow PRO with different membrane area were carried
out and the results are presented in Fig. 4. As in the previous figures,
the solid lines are for draw solution and dashed lines for feed solution in
Fig. 4a and b. It is surprising to find that the membrane areas smaller
than 400 m2 are insufficient for equilibrium to be reached in the
membrane channels. For the larger membrane areas (500 m2 and
600 m2), the flat portions in Fig. 4a and b and the region of zero flux in
Fig. 4c adjacent to the draw entrance indicate that equilibrium is
45
Desalination 433 (2018) 41–47
a
Fig. 3. (a) Flow rates, (b) salt concentrations, and (c) permeate flux in counter-current
flow PRO with feed fraction θ = 0.8. The solid curves in (a) and (b) represent the draw
solution and the dashed curves represent the feed solution. Numbers 1, 2, 3, 4 indicate the
total membrane area of 20 m2, 40 m2, 80 m2, and 100 m2, respectively.
reached in the PRO systems. At equilibrium, the salt concentrations of
feed and draw solutions leaving the membrane channels are cF∗ and cD∗,
respectively. It can be found that the minimal total membrane area
required to reach equilibrium is roughly 443 m2, which is much greater
than those in the cases for θ = 0.2 and 0.8. For convenient comparison,
the membrane areas required to reach equilibrium at different feed
fractions are identified from numerical solutions and presented in
Table 2.
Distributions of water flux along the channel are presented in
Fig. 4c. For the smallest membrane area of 100 m2, water flux increases
monotonically from the feed entrance to the draw entrance. For
membrane areas of 200 m2, 300 m2, and 400 m2, water flux starts from
a nonzero value to a peak value at a point near the draw entrance and
then drops sharply to another nonzero value. For membrane areas of
500 m2 and 600 m2, because equilibrium has been reached in the PRO
channels, water flux starts from zero at feed entrance, increases to a
peak value at about 430 m2, and drops abruptly to zero at about
443 m2. The remaining length of the membrane channel to the draw
entrance is a dead region of zero water flux. The flux curves for the
membranes areas of 500 m2 and 600 m2 are identical before the dead
region. It is interesting to find that the shape of water flux curve for
critical feed fraction at equilibrium resembles the left part (decreasing
W. Yang et al. Desalination 433 (2018) 41–47

a a

b b

c Fig. 5. (a) Power density, and (b) specific energy in counter-current flow PRO are pre-
sented as function of total membrane area. Numbers 1, 2, 3 indicate feed fraction 0.2,
0.426, and 0.8, respectively.

∆P ∆Q
PD =
S (18)
and
∆P ∆Q
SE =
QF 0 + QDS (19)
where PD is the power density, SE is the specific energy, ΔQ is the
transmembrane water flow rate in the entire module. The transmem-
brane water flow rate is readily determined from the numerical solution
Fig. 4. (a) Flow rates, (b) salt concentrations, and (c) water fluxes in counter-current flow as Δ Q = QF0 − QFS or Δ Q = QD0 − QDS.
PRO with the critical feed fraction θ⁎ = 0.426. The solid curves represent the draw so- Power density and specific energy are calculated for feed fraction
lution and the dashed curves represent the feed solution. The numbers 1, 2, 3, 4, 5, 6 0.2, 0.426 and 0.8 as functions of the total membrane area and the
represent the total membrane area of 100 m2, 200 m2, 300 m2, 400 m2, 500 m2, and
results are presented in Fig. 5. It can be seen from Fig. 5a that feed
600 m2, respectively.
fraction has enormous impact on power density. Power density is
higher for smaller feed fraction because higher salinity gradient can be
Table 2 maintained. Power density decreases monotonously with the increasing
Membrane area needed for equilibrium (m2).
total membrane area due to decrease of the average salinity gradient.
θ 0.1 0.2 0.3 0.4 0.426 0.5 0.6 0.7 0.8 0.9 Fig. 5b shows that specific energy behaves different from power density
and the impact of feed fraction on specific energy is mixed. Portions of
S 64 75 98 184 443 265 174 111 64 28 constant specific energy on curves 1 and 3 in Fig. 5b indicates that
equilibrium is reached in these PRO processes. Specific energy for feed
fraction of 0.8 starts with the highest initial flux at membrane area of
gradually to zero toward feed entrance) in case of larger feed fraction
10 m2 but reaches soonest to its maximum value of 0.058 kWh/m3 at
θ > θ⁎ and the right part (reaching the peak value and abruptly
membrane area of about 40 m2. Specific energy for feed fraction of 0.2
dropping to zero before a dead region of zero flux to the draw entrance)
started with the lowest initial value and reaches to its maximum value
of the case of smaller feed fraction θ < θ⁎ at equilibria.
of 0.078 kWh/m3 at membrane area of about 75 m2. The curve for feed
fraction of 0.426 represents the highest specific energy except for a
3.3. Power density and specific energy small section at the very beginning. Specific energy increases mono-
tonically to 0.13 kWh/m3 at the largest total membrane area shown in
As one of the potential technologies to harness salinity gradient this figure but equilibrium is not reached.
energy, PRO process is eventually assessed in terms of power genera- The impacts of feed fraction on power density and specific energy
tion. Two metrics are commonly used to quantify the performance of for four different total membrane areas are shown in Fig. 6. As in-
power generation. One is the membrane power density defined as the dicated before, the feed flow rate is kept constant during numerical
power generated per membrane area. The other is the specific energy simulation, while the draw flow rate decreases as the feed fraction in-
defined as the energy extracted per total volume of the feed and draw creases. It can be seen from Fig. 6a that power density in general de-
solutions combined. The average power density and specific energy are, creases with the increasing feed flow rate fraction. The flat portion of
respectively, determined by curve 4 may indicate that the availability of feed water in is the limiting

46
W. Yang et al.
a distribution of water flux along the membrane channels and the loca-
b References
Fig. 6. (a) Power density, and (b) specific energy in counter-current flow PRO are pre-
sented as function of feed fraction. Numbers 1, 2, 3, 4 indicate membrane area of 20 m2,
40 m2, 60 m2, 80 m2, respectively.
factor. Under this condition, the recovery of the feed solution is not
affected much by the amount of draw solution. Specific energy for all
cases increases initially with feed fraction to a peak value and then
decreases with further increasing feed fraction. It is similar to the be-
havior of the Gibbs free energy of mixing feed and draw solutions re-
ported in the literate [18]. With the same feed fraction, power density
decreases with increasing membrane area while specific energy does
the opposite.
4. Conclusion
A systemic numerical procedure has been developed in this study
for rigorously modeling and simulation of the counter-current flow PRO
system. It is demonstrated that the procedure works well no matter
equilibrium is established in the system or not. The change of flow rates
and salt concentrations along the membrane channels for any feed
fractions can be easily obtained with the numerical procedure. These
numerical solutions can provide detailed insight information into the
process that are critical to understand the performance and to improve
the design of PRO system. Particularly, the knowledge about the
47
Desalination 433 (2018) 41–47
tion for the zero flux region at various feed fractions is very useful to
optimize PRO design and operation. The average power density and
specific energy of the PRO process, the two commonly used parameters
for assessment of energy production efficiency, are readily calculated
from the numerical solutions.
Acknowledgments
This work was supported in part by China Scholarship Council, in
part by Fundamental Research Funds for the Central Universities
(CHD2011JC114).
[1] R.E. Pattle, Production of electric power by mixing fresh and salt water in the hy-
droelectric pile, Nature 174 (1954) 660.
[2] R.S. Norman, Water salination: a source of energy, Science 4161 (1974) 350–352.
[3] S. Loeb, Osmotic power plants, Science 4203 (1975) 654–655.
[4] B.E. Logan, M. Elimeleth, Membrane-based processes for sustainable power gen-
eration using water, Nature 488 (2012) 313–319.
[5] G.L. Wick, Power from salinity gradients, Energy 3 (1978) 95–100.
[6] N.Y. Yip, M. Elimelech, Thermodynamic and energy efficiency analysis of power
generation from natural salinity gradients by pressure retarded osmosis, Environ.
Sci. Technol. 46 (2012) 5230–5239.
[7] K.P. Lee, T.C. Arnot, D. Mattia, A review of reverse osmosis membrane materials for
desalination–development to date and future potential, J. Membr. Sci. 370 (2011)
1–22.
[8] S. Chou, R. Wang, L. Shi, Q.H. She, C.Y. Tang, A.G. Fane, Thin-film composite
hollow fiber membranes for pressure retarded osmosis (PRO) process with high
power density, J. Membr. Sci. 389 (2012) 25–33.
[9] J.W. Post, H.V.M. Hamelers, C.J.N. Buisman, Energy recovery from controlled
mixing salt and fresh water with a reverse electrodialysis system, Environ. Sci.
Technol. 42 (2008) 5785–5790.
[10] J. Veerman, M. Saakes, S.J. Metz, G.J. Harmsen, Electrical power from sea and river
water by reverse electrodialysis: a first step from the laboratory to a real power
plant, Environ. Sci. Technol. 44 (2010) 9207–9212.
[11] G.Z. Ramon, B.J. Feinberg, E.M.V. Hoek, Membrane-based production of salinity-
gradient power, Energy Environ. Sci. 4 (2011) 4423–4434.
[12] A. Achilli, A.E. Childress, Pressure retarded osmosis: from the vision of Sidney Loeb
to the first prototype installation — review, Desalination 261 (2010) 205–211.
[13] K. Nijmeijer, S. Metz, Chapter 5 salinity gradient energy, Sustain. Sci. Eng. 2 (2010)
95–139.
[14] T. Thorsen, T. Holt, The potential for power production from salinity gradients by
pressure retarded osmosis, J. Membr. Sci. 335 (2009) 103–110.
[15] A. Achilli, T.Y. Cath, A.E. Childress, Selection of inorganic-based draw solutions for
forward osmosis applications, J. Membr. Sci. 364 (2010) 233–241.
[16] Statkraft, The world's first osmotic power prototype opens today. November 2009,
http://www.statkraft.com/media/press-releases/Press-releases-archive/2009/ ,
Accessed date: 16 December 2017.
[17] Statkraft, Statkraft halts osmotic power investments. December 2013, http://www.
statkraft.com/media/news/News-archive/2013/Statkraft-halts-osmotic-power-
investments , Accessed date: 16 December 2017.
[18] S. Lin, A.P. Straub, M. Elimelech, Thermodynamic limits of extractable energy by
pressure retarded osmosis, Energy Environ. Sci. 7 (2014) 2706–2714.
[19] A.P. Straub, S. Lin, M. Elimelech, Module-scale analysis of pressure retarded os-
mosis: performance limitations and implications for full-scale operation, Environ.
Sci. Technol. 48 (2014) 12435–12444.
[20] A.P. Straub, S. Lin, M. Elimelech, Pressure-retarded osmosis for power generation
from salinity gradients: is it viable? Energy Environ. Sci. 9 (2016) 31–48.