Desalination 447 (2018) 133–146

Contents lists available at ScienceDirect

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

Experimental and simulation studies of two types of 5-inch scale hollow T

fiber membrane modules for pressure-retarded osmosis
Yasuhiko Tanakaa, Masahiro Yasukawaa,b, Shohei Godac, Hidehiko Sakuraic, Masafumi Shibuyaa,
⁎ ⁎⁎
Tomoki Takahashia, Michimasa Kishimotoa, Mitsuru Higaa,b, , Hideto Matsuyamaa,
a
Center for Membrane and Film Technology, Department of Chemical Science and Engineering, Kobe University, 1-1 Rokkodaicho, Nada-ku, Kobe 657-8501, Japan
b
Division of Applied Chemistry, Graduate School of Sciences and Technology for Innovation, Yamaguchi University, 2-16-1, Tokiwadai, Ube, Yamaguchi 755-8611, Japan
c
Toyobo Co. Ltd., 2-8 Dojima Hama, 2-chome, Kita-ku, Osaka 530-8230, Japan

G R A P H I C A L A B S T R A C T

A R T I C LE I N FO A B S T R A C T

Keywords: This study experimentally and theoretically analyzed the performance of two types of large-scale hollow fiber
Pressure-retarded osmosis (HF) forward osmosis (FO) modules for pressure retarded osmosis (PRO). The effects of operating conditions on
Cellulose triacetate membrane the module performance of the 5-inch scale HF modules with a cross-wound HF configuration were investigated.
Hollow fiber membrane module A modified analytical model, based on the friction-concentration polarization (FCP) model, which combined the
Friction-concentration polarization model
PRO theory with water flux and salt leakage, was proposed for PRO performance estimation. The theoretical
results agreed within 9.7% deviation with the experimental results under all conditions. The energy efficiency of
the HF PRO module was also theoretically derived. The power generation estimation for the 5-inch membrane
module revealed that 10 to 15% of the energy could be recovered from the reverse osmosis seawater desalination
process. However, some parts of the membrane could not be used efficiently inside the modules because of the
non-optimal dimensions. Therefore, new types of modules, having shorter lengths and larger module diameters,
were proposed and provided greater net energy output, as compared with the original module, due to the
reduction of both the region where the water was not sufficiently permeated and the pressure drop inside the HF
membrane.

⁎
Correspondence to: M. Higa, Center for Membrane and Film Technology, Department of Chemical Science and Engineering, Kobe University, 1-1 Rokkodaicho,
Nada-ku, Kobe 657-8501, Japan.
⁎⁎
Corresponding author.
E-mail addresses: mhiga@yamaguchi-u.ac.jp (M. Higa), matuyama@kobe-u.ac.jp (H. Matsuyama).

https://doi.org/10.1016/j.desal.2018.09.015
Received 27 December 2017; Received in revised form 18 September 2018; Accepted 18 September 2018
0011-9164/ © 2018 Published by Elsevier B.V.
Y. Tanaka et al. Desalination 447 (2018) 133–146

1. Introduction
Reverse osmosis (RO) is widely used as a seawater desalination
technique due to its lower energy cost than the previously conventional
evaporation method, owing to the absence of latent heat energy. In
recent years, the energy consumption has been further reduced by in-
cluding a pressure exchanger that enables the recovery of the pumping
energy from RO desalination [1]. Recent RO developments have
achieved a low energy consumption of approximately 2.5 kWh·m−3 for
water production [2,3], approaching the theoretical minimum value of
approximately 1.0 kWh·m−3 [4]. However, further reduction of the
energy consumption is still required, due to the growing demand for
secure water resources. In addition, the large amount of concentrated
brine, an RO desalination plant by-product, that is disposed into the
surrounding environment should also be considered [5,6].
Accordingly, pressure retarded osmosis (PRO) has attracted atten-
tion as an emerging technology to solve these problems [5,7–9]. PRO
enables the conversion of the salinity gradient energy into electric en-
ergy by using a semipermeable membrane [10–13]. When a PRO
system is combined with a modern RO desalination plant, approxi-
mately 3.87 MJ·m−3 of energy can potentially be generated due to the
Gibbs free energy of mixing between the RO brine (with an assumed
concentration of 1.2 mol L−1) and another low-salinity water, such as
wastewater or treated sewage (with an assumed concentration of
10 mmol L−1) [14]. In addition, since the PRO process also enables the
dilution of the concentrated seawater, the environmental impact of the
disposed RO brine can be sufficiently reduced [15].
Recently, in order to investigate and improve the performance of
PRO, many studies have been conducted at a lab scale [16–23]. How-
ever, to estimate the feasibility, especially of a PRO pilot plant, a pilot
scale experimental study using the membrane module is required in
addition to the lab scale experiment because the PRO performance at
the pilot scale is highly influenced by not only the PRO membrane
performance, which is considerable at the lab scale, but also the oper-
calculations, we also modified the FCP model by combining the PRO
theory for water flux and salt leakage with our previous model [28].
Furthermore, the energy efficiency of the HF PRO module was theo-
retically obtained by comparing its performance with the Gibbs free
energy of mixing under different operating conditions.
2. Theory
2.1. Modified FCP model for PRO
The friction concentration polarization (FCP) model [29] is an in-
itial analytical model of an HF RO module in which the pressure loss
inside the HF is considered based on the Kimura-Sourirajan membrane
transport equation [30]. The FCP model is based on the stepwise seg-
mental (integral) calculation of both the inner bore and outer shell sides
of the HF membrane, considering the external concentration polariza-
tion, pressure drops, and concentration profiles in the RO module. In
our previous study [28], we combined FO theory with an FCP model in
order to estimate the HF module performance under FO operation.
Herein, to apply this modified FCP model for PRO operation, we have
improved the model by combining PRO theory with the FCP model.
Osmotically driven water flux is often expressed using the intrinsic
parameters of the membrane, such as the water permeability coeffi-
cient, A [L·m−2·h−1·bar−1], solute permeability coefficient, B
[L·m−2·h−1], and structural parameter, S [μm]. In the case of PRO
operation, the water flux, Jw [L·m−2·h−1], can be expressed by con-
sidering the membrane orientation and the internal and external con-
centration polarization effects, based on the solution diffusion model, as
follows [17]:
Jw = A·
⎧π
( ) Jw
( )
Jw S
⎪ DS, b exp − k − πFS, b exp Ddiff
− ΔPDS − FS
⎫
⎪
⎨ ⎬
⎩
B ⎡

⎣
( )
Jw S
⎪ 1 + Jw exp Ddiff − exp − k ( )
Jw ⎤

⎦
⎪
⎭ (1)
ating conditions, module characteristics, and module dimensions. To
where πDS,b [bar] and πFS,b [bar] are the osmotic pressures in the bulks
date, although there have been a few pilot scale studies of PRO process
of the more concentrated (draw solution: DS) and less concentrated
performance [24–26], further experimental investigation of the PRO
solutions (feed solution: FS), respectively; ΔPDS−FS [bar] denotes the
process on a pilot scale is still required. Moreover, accurate models for
hydraulic pressure difference between the shell side DS and the bore
estimating the performance of PRO modules at the pilot scale are still
side FS of the HF membrane; and Ddiff [μL·m−1·h−1] and k [L·m−2·h−1]
lacking, and, therefore, optimized module-specific models should be
are the diffusivity and mass transfer coefficient, respectively, of the
developed. Thus, both experimental and theoretical studies using pilot
draw solute.
scale PRO membrane modules are required for estimating, designing,
The reverse solute flux (Js [mol·m−2·h−1]) from the DS to FS can be
and improving pilot scale PRO systems.
expressed as follows [31]:
Recently, several types of module configurations, such as hollow
fiber (HF) [24], spiral wound [25,26], and plate and frame [27] con- Js =
Jw + AΔPDS − FS
A
figurations have been proposed for pilot scale PRO membrane modules. ·νRT (2)
B
Among them, the HF module has been proposed as an interesting ele-
−1 −1
ment configuration because of its high packing density and the appro- where ν [−], R [bar·L·mol ·K ], and T [K] are the van't Hoff factor,
priate flow patterns in the module [24,28]. In addition, current HF RO gas constant, and solution temperature, respectively.
modules can be directly used as PRO modules because they have 4 In the FCP model, the pressure drop at the shell side of the HF (DS
ports, including an inner HF bore side inlet and outer HF shell side side) is expressed by the Ergun equation as follows [29,32]:
outlet ports, without additional modification required for the spiral dPshell 150(1 − ϵ)2 μshell ushell 1.75(1 − ϵ) ρshell ushell 2
wound membrane module [28]. = 3
· 2
+ ·
dr ϵ (1.5dshell ) ϵ3 1.5dshell (3)
In our previous study, we experimentally and theoretically in-
vestigated osmotically driven module performance using a 5-inch scale where ε [−] is the porosity of the HF module; μshell [Pa·s], ushell
HF membrane module under forward osmosis (FO) operation [28]. A [m·s−1], and ρshell [kg·m−3] are the viscosity, flow rate, and density of
simple integral calculation model, developed from the friction con- the outer shell side solution of the HF membrane, respectively; and r
centration polarization (FCP) model, was proposed in order to theore- [m] is the coordinate in the radial direction.
tically explain and predict the HF module performance under FO op- The pressure drop at the inner bore side of the HF membrane (FS
eration. Although our developed model agreed well with the side) is expressed by the Hagen-Poiseuille equation, as follows
experimental data for an HF module under a wide range of FO oper- [28,29,32]:
ating conditions, its versatility is still unclear, especially for different 32μbore ubore
dPbore
HF modules under PRO operation. =
dz din 2 (4)
In this study, we experimentally and theoretically investigated the
−1
pilot scale performance of two types of HF modules under a wide range where din [m], z [m], μbore [Pa·s], and ubore [m·s ] are the inner dia-
of PRO operating conditions. To improve the accuracy of the meter of the HF, axis coordinate, bore side solution viscosity, and bore

134
Y. Tanaka et al. Desalination 447 (2018) 133–146

side solution flow rate, respectively. To predict the pressure drop more respectively; and α, β, and γ are empirical parameters, which can be
accurately, we also considered the difference in the HF length in the determined from experimental data fitting [28]. In our case, because Sc
radial direction because the length of HF wound in the module depends is a parameter that depends on the kind of solution, and we used only
on the winding radius and angle [28]. The length of HF in the module NaCl solution in this study, γ was fixed to be 0.33 based on the Deissler
can be calculated by using wound number, W [−], of fibers around a equation. On the other hand, α and β are empirical parameters, and
central core tube as shown in the previous literatures [29,32]. their values depend on only the module characteristics, such as the
In the case of the module experiment, the changes in the osmotic module dimensions, flow pattern, and so on. Hence, we determined α
driving force due to the DS dilution, FS up-concentration, and con- and β for the respective modules by minimization of the residual sum of
centration distribution in the module cannot be ignored. Therefore, to squares (RSS) as follows [34]:
consider them, a stepwise segmental (integral) calculation was per-
formed. The integral calculation was conducted with 50,000 segmental
(RSS ) = ∑ {(Jwexp . − Jwcalc.)2 + (Jsexp . − Jscalc.)2} (13)
cells (100 and 500 in the radial (r) and axial (z-axis) directions, re- where Jw exp
and Js are the experimental flux and reverse salt flux
exp

spectively). We preliminarily confirmed that the number of segmental data for the HF module, respectively, and Jwcalc and Jscalc are the the-
cells, 50,000, was sufficient for the accurate calculation of the HF oretical values calculated according to the methodology described
module performance in this study as shown in the Supplementary above. The detailed procedure is shown in the Supplementary In-
Information section. The concentrations of the DS at the (z(n), r formation section.
(m + 1)) segment and the FS at the (z(n + 1), r(m)) segment in the HF
module are expressed as follows. 2.2. Extractive energy by PRO
CDS (z (n), r (m)) ·QDS (z (n), r (m)) − Js (z (n), r (m)) ·Am (z (n), r (m))
CDS (z (n), r (m + 1)) =
QDS (z (n), r (m)) + Jw (z (n), r (m)) ·Am (z (n), r (m)) To determine the PRO module efficiency, we compared the PRO
power output calculated from the experimental data with the theore-
(5)
tical maximum extractable work, Wp [J], calculated using the metho-
CFS (z (n), r (m)) ·QFS (z (n), r (m)) + Js (z (n), r (m)) ·Am (z (n), r (m)) dology in the literature [14]. To calculate Wp for a PRO process using a
CFS (z (n + 1), r (m)) =
QFS (z (n), r (m)) − Jw (z (n), r (m)) ·Am (z (n), r (m)) perfect PRO membrane, the maximum applied hydraulic pressure was
(6) first calculated using the permeated water volume, ΔV [m3], as follows
−1 −1
[14]:
where CDS [mol·L ] and CFS [mol·L ] are the solute concentrations of
the shell side DS and bore side FS, respectively; QDS [L·h−1] and QFS V0 V0
ΔPDS − FS = ΔπDS − FS = νRT ⎛⎜ 0 DS 0
CDS − 0 F 0 ⎞
CFS
[L·h−1] are the flow rates of the shell and bore sides, respectively, of the
⎟

⎝ V DS + ΔV V FS − ΔV ⎠ (14)
HF membrane; and Am(z(n), r(m)) [m2] is the effective HF membrane area
at the segmental location (z(n), r(m)). The flow rate at the next adjacent where ΔPDS−FS [bar] and ΔπDS−FS [bar] are the hydraulic and osmotic
segment, (z(n), r(m + 1)) and (z(n + 1), r(m)), was derived by material pressure differences, respectively, between the DS and FS; V0 [m3] and
balance, as follows: C0 [mol·m−3] are the volume and concentration of the initial solution;
and ν [−], R [bar·m3·mol−1·K−1], and T [K] are the van't Hoff factor,
QDS (z (n), r (m + 1)) = QDS (z (n), r (m)) + Jw(z (n), r (m)) ·Am (z (n), r (m)) (7) gas constant, and absolute temperature of the solution, respectively.
QFS (z (n + 1), r (m)) = QFS (z (n), z (m)) − Jw(z (n), r (m)) ·Am (z (n), r (m)) The work generated by a PRO operation under a constant ΔPDS−FS per
(8)
unit volume of DS at the inlet, EDS, intheo. [J·m−3-DS], can then be cal-
The integral scheme for calculating the PRO HF module perfor- culated using the ΔV value, as follows:
mance is shown in Fig. 1. First, the set conditions (CDS,in, CFS,in, and
V0 V0
PDS,out) were selected, and Eqs. (1)–(8) were then calculated for the Wp = ΔPDS − FS ΔV = νRT ⎛⎜ 0 DS 0
CDS − 0 FS 0 ⎞
CFS ⎟ ΔV

section from m = 1 to m = 100, along the radial direction under n = 1. ⎝ V DS + ΔV V FS − ΔV ⎠ (15)

This calculation was subsequently performed for n = n + 1 to n = 500.
theo .
Wp
Before the calculation, the bore side FS inlet pressure, PFS,in, was as- EDS , in = 0
V DS (16)
sumed to be a certain value. The applied inlet DS pressure, PDS,in, was
changed in each calculation. After the calculation for n = 500 and This theoretical EDS, intheo. was compared with the net energy output
m = 100 was complete, we fixed the PFS,in, unless the FS outlet pres- per unit volume of DS of the PRO system, EDS, innet, calculated as shown
sure, PFS,out, was nearly equal to zero. Because the FS outlet was open to in Section 3.4.
the atmosphere, PFS,out should be zero. When the set PFS,in satisfied the
condition of PFS,out = 0, the calculation was finished. 3. Experimental
In the FCP model [29], the mass transfer coefficient, k, is expressed
by non-dimensional equations as follows [29,32,33]: 3.1. Materials and chemicals
Sh = α·Re β ·Sc γ (9)
For the membrane module performance tests, analytical grade so-
Ddiff ⎞ β γ dium chloride (Wako Pure Chemical Co., Osaka, Japan) was used to
k = α⎛ ⎜ Re ·Sc
⎟

⎝ dout ⎠ (10) prepare the salt solution used as the DS. After purification with an ac-
tivated carbon filter, tap water was used as the FS and as the DS solvent.
dout ushell ρshell
Re =
μshell (11) 3.2. HF membrane modules
μshell
Sc = Two types of 5-inch scale HF membrane modules were kindly pro-
ρshell Ddiff (12)
vided by TOYOBO Co. Ltd., Japan. A schematic illustration of the HF
where Sh [−], Re [−], and Sc [−] are the Sherwood, Reynolds, and module is shown in Fig. 2. In HF modules, the HFs are cross-wound
Schmidt numbers, respectively; Ddiff [m2·s−1] and dout [m] are the dif- around a central tube. The DS flows between the HF bundles from the
fusion coefficient of the draw solute and the outer diameter of the HF, central tube in the direction radial to the flow. Both ends of the fibers
respectively; ushell [m·s−1], ρshell [kg·m−3], and μshell [Pa·s] represent are open, and the FS flows into the bore side of the HF module. The HFs
the flow rate, density, and viscosity of the shell side solution, have outer active layers, and therefore, the osmotically driven water

135
Y. Tanaka et al. Desalination 447 (2018) 133–146

Fig. 1. Calculation scheme for the PRO HF module performance.

Fig. 2. Schematic illustration of a PRO HF module with a cross-wound HF configuration.

136
Y. Tanaka et al. Desalination 447 (2018) 133–146

Table 1
Technical specifications of the HF membrane modules.
HF inner diameter, din HF outer diameter, dout Module length Number of fibers Effective membrane area, Packing density Wound number
[μm] [μm] [mm] [−] Am [m2] [%] [−]

Module (1) 85 175 682 220,000 70.5 54.2 2

Module (2) 117 189 682 165,600 65.6 43.2 2

Table 2 regeneration to provide constant DS and FS inlet quality. The DS and FS

Intrinsic membrane parameters of the HF membranes within the modules.
Water permeability Salt permeability
coefficient, A coefficient, B
[L·m−2·h−1·bar−1] [L·m−2·h−1]
Module (1) 0.27 0.035
Module (2) 0.60 0.072
flows from the inner bore side to the outer shell side of the HF mem-
brane under PRO operation.
In this study, we used two kinds of modules made of two different
cellulose triacetate (CTA) HF membranes. Table 1 presents the HF
membrane and module dimensional information. The two types of HF
membranes had different permeabilities and structures (cf. inner and
outer diameters). Module (1) had HFs of smaller inner and outer dia-
meters than those in module (2). Because of the difference in the outer
diameters, the number of HFs and the resulting effective membrane
areas within the modules were different. The water permeability (A-
value), salt permeability (B-value), and structure parameter (S-value) of
the HF membranes were preliminarily measured using RO and FO tests
that employed a mini-module, as described in our previous studies
[35,36]. The mini-module experiments are preferable to estimate the
membrane intrinsic characteristics because we can eliminate various
effects such as pressure drop, DS dilution, FS up-concentration and
external concentration polarization that should be considered in a pilot
scale module analysis. The A-, B-, and S-values obtained for the two
types of HF membranes are shown in Table 2. As compared with those
in module (1), the HFs installed in module (2) had higher A- and B-
values and a similar S-value.
3.3. PRO tests
The PRO tests were carried out using the PRO/RO hybrid system
described in our previous reports [24,28]. A schematic illustration of
this system is shown in Fig. 3. The RO system was used for DS
were 1.0 M NaCl and tap water, respectively, and were introduced to
Structural the outer shell and inner bore sides, respectively, of the HF membrane
parameter, S in the module. The applied pressure was carefully adjusted from 5 to
[μm] 30 bar and 3 to 12 bar for module (1) and module (2), respectively. The
lower applied hydraulic pressure range for module (2) was due to the
1024
912
pressure resistance limitation of its HFs. The flow rates (QDS,in, QFS,in,
QDS,out, and QFS,out), pressures (PDS,in, PDS,out, and PFS,in), conductivities
(CDS.in, CFS,in, CDS,out, and CFS,out), and temperatures (TDS,in and TFS,in)
were monitored using a multichannel logging system connected to a
personal computer. The DS and FS inlet flow rates were controlled to
8–14 L·min−1. The average water flux was calculated from the experi-
mental data using the following equation:
QFS, in − QFS, out
Jw =
Am (17)
where QFS,in [L·h−1] and QFS,out [L·h−1] are the inlet and outlet FS flow
rates, respectively, and Am [m2] is the effective membrane area of the
module.
3.4. Estimation of extractive net work per DS volume from the PRO
experiment
To estimate the energy efficiency of the PRO system, we simulated
the extractive work calculated from the PRO experiment, EDS, innet,
under various operating conditions. Fig. 4 shows a simple flow diagram
for a PRO system consisting of a PRO module, pumps, a pressure ex-
changer (PX) for energy recovery from the pump, and a turbine for
power generation. To simply estimate the system performance due only
to the module, the efficiencies of the turbine and pressure exchanger
were assumed to be 1, and it was also assumed that the permeated
water, QJw, was used for the power generation, and QDS,in in all QDS,out
returned to the PX for energy recovery. The net power output of the
PRO system, Enet, was then estimated by subtraction of the DS and FS
pumping power from the generated power, Eturbine, as estimated from
the water flux of the modules as follows:

Fig. 3. Schematic illustration of the PRO/RO hybrid system. E: conductivity meter, F: flow meter, P: pressure gauge, T: temperature meter, CP: circulation pump, LP:
low-pressure pump, HP: high-pressure pump, CF: cartridge filter, FP: feed pump.

137
Y. Tanaka et al.
Fig. 4. Flow diagram of the PRO system evaluated in this study.
pump pump
E net = E turbine − (EDS + EFS ) FS up-concentration at the inner bore side of the HF membrane due to
net
E = QJw PDS, out − (QDS, in (PDS, in − PDS, out ) + QFS, in PFS, in)
pump
where EDS and EFSpump
are the powers consumed by the high-pres-
sure DS pump and low-pressure FS pump, respectively.
For an RO desalination process, the minimum energy required to
obtain 1 m3 of pure water for a recovery ratio of 50% can be theore-
tically calculated as 3.78 MJ·m−3-pure water (1.05 kWh·m−3-pure
water) [2]. Recently developed real RO processes require approxi-
mately 9.0 MJ·m−3-pure water (2.5 kWh·m−3-pure water) [3]. Because
a 50% recovery ratio is assumed, these values per 1 m3 of pure water
are the same as those per 1 m3 of brine. To estimate the efficiency and
feasibility of the PRO system more quantitatively, by comparing to the
energy consumption of an RO desalination process, Enet was normalized
by the DS flow rate as follows:
E net
net
EDS , in =
QDS, in
where EDS, innet is the extractive net work per DS volume obtained from
the PRO experiment. To express the inlet flow ratio of QDS,in, ΦDS,in was
also defined as follows:
QDS, in
ΦDS, in =
QDS, in + QFS, in
Theoretically, the power generated in the PRO process reaches its
maximum when the applied hydraulic pressure is half of the osmotic
pressure of the DS [13,37,38], however, the optimal applied pressure
for the maximum EDS, innet depends on the experimental conditions, such
as the inlet flow ratio, membrane performance, and module char-
acteristics, when considering the DS dilution, FS up-concentration, salt
leakage, and pressure drops within the module. To determine the op-
timum operating pressure for the respective HF modules, we estimated
EDS, innet as a function of PDS.in with different inlet flow rates at first.
After then, we also compared EDS, innet to the theoretical maximum
energy output of the PRO system for different flow rates under a con-
stant PDS,in.
4. Results and discussion
4.1. Effects of operating conditions on PRO performance
Fig. 5(a) shows the water flux, Jw [L·m−2·h−1], as a function of the
applied inlet pressure, PDS,in [bar] for different inlet flow rate condi-
tions using module (1) (Fig. 5(a–1)) and module (2) (Fig. 5(a–2)). These
data with the relative errors were also shown in the Supplementary
Information. Because of the pressure resistance limitation of the HFs in
the modules, the applied hydraulic pressures for module (1) and
module (2) were < 30 and 12 bar, respectively. As shown in these fig-
ures, when the applied hydraulic pressure increased, Jw decreased due
to the decrease in the driving force. In both modules, Jw increased with
increasing QDS,in because the increase in QDS,in reduced both the DS
dilution effect due to the water permeation and the external con-
centration polarization (ECP) effect at the outer shell side of the HF
membrane [28]. Jw also increased with increasing QFS,in because of the
Desalination 447 (2018) 133–146
(18) the water permeation being reduced. In the case of module (2), when
QFS,in was less than approximately 10 L·min−1 and the applied pressure
was less than approximately 12 bar, Jw became constant, as shown in
Fig. 5(a–2), because all the FS was permeated [28]. When all the QFS,in
was not permeated, module (2) had water flux approximately two times
greater than that of module (1) because of the higher water perme-
ability of the HF membrane, as shown in Table 2.
Fig. 5(b) shows the FS inlet pressure, PFS,in [bar], as a function of
PDS,in for different inlet flow rate conditions using module (1)
(Fig. 5(b−1)) and module (2) (Fig. 5(b−2)). Because the FS outlet was
open to the atmosphere, PFS,out was almost equal to the atmospheric
pressure. PFS,in increased with increasing QFS,in due to the increased
pressure drop and decreased with decreasing PDS,in because QFS de-
creased due to the increase of the permeated water flux from the FS to
DS. When QDS.in increased, PFS,in also decreased due to the increase in
the permeated water flux. PFS.in for module (1) was higher than that for
(19)
module (2) because the HFs within module (1) had smaller din (85 μm)
than those within module (2) (117 μm) and demonstrated a more severe
pressure drop.
For the theoretical calculation, we experimentally determined the
empirical values of α and β in Eqs. (9)–(10) in order to express the ECP
effects within the respective HF modules. By minimizing the RSS of Eq.
(20)
(13), α and β were set at 0.45 and 0.10, respectively, for module (1) and
0.60 and 0.05, respectively, for module (2). The detailed RSS data for
the HF modules is shown in the Supplementary Information. The the-
oretical calculation results for Jw and PFS,in are also shown in Fig. 5 as
solid lines. All the calculated results were in good agreement with the
experimental data (9.6% and 6.5% of the maximum deviation from the
experimental data for modules (1) and (2), respectively) under various
operating conditions in both modules (1) and (2).
4.2. Estimation of net energy output per DS volume from the PRO under
desirable PDS,in
Because the theoretical simulations for modules (1) and (2) agreed
well with all experimental data, as shown in Fig. 5, it enabled to si-
mulate EDS, innet under various operating conditions. Fig. 6 shows the
estimated EDS, innet as a function of PDS.in with different inlet flow rates.
As shown in the literature [14], the optimum PDS.in for the maximum
EDS, innet depends on the ΦDS,in and decreases with increasing ΦDS,in
because πDS,out at the equilibrium state increases with increasing ΦDS,in,
allowing for a higher PDS.in. In the case of module (1), because of its
higher pressure-resistance (< 30 bar), as mentioned in Section 3.3, EDS,
in
net
could reach its maximum when ΦDS was less than approximately
0.7, whereas, in the case of module (2), because of the lower pressure-
resistance (< 10 bar) of its HFs, the EDS, innet did not reach its maximum.
Higher pressure would be necessary to obtain the maximum energy
from module (2).
Although the EDS, innet of module (2) was greater than that of module
(1) for QDS,in < 4 L·min−1, the EDS, innet of module (1) became greater
than that of module (2) when QDS,in > 4 L·min−1. Consequently, these
theoretical calculations indicate that approximately 0.8–1.9 MJ·m−3-
pure water (20–50% of the theoretical minimum energy and 9–21% of
138
Y. Tanaka et al.
Fig. 5. Effect of the inlet pressure applied on the shell side, PDS,in, on (a) PRO water flux and (b) FS inlet pressure, PFS,in. The units of QDS,in and QFS,in are all L·min−1.
Solid lines denote the calculated results.
the actual energy consumed by RO desalination) and 0.5–2.4 MJ·m−3-
pure water (13–63% of the theoretical minimum energy and 6–26% of
the actual energy consumed by RO desalination) could potentially be
recovered under a desirable PDS,in condition by using modules (1) and
(2), respectively.
4.3. Estimation of net PRO power output per DS volume under constant
PDS,in
In order to further evaluate the PRO performance, we also compared
EDS, innet to EDS, intheo. for different flow rates under a constant PDS,in.
Hereafter, the PDS,in was fixed at 20 and 10 bar for modules (1) and (2),
respectively, because of their pressure tolerance limits. Fig. 7 shows the
relationship between the inlet flow rates (QDS,in and QFS,in) and EDS, innet.
The theoretical maximum extractive work of the PRO, EDS, intheo, as a
function of ΦDS,in for a constant PDS,in is also shown in Fig. 7. These
theoretical values were calculated using Eq. (16). The EDS, innet de-
creased with increasing QDS,in in both modules (1) and (2). The op-
timum QFS,in increased with increasing QDS,in. Compared to the theo-
retical maximum PRO power output for a constant PDS,in (dotted line in
Fig. 7), EDS, innet was nearly equal to the EDS, intheo for ΦDS,in greater than
approximately 0.5. However, these deviated from the EDS, intheo with
decreasing ΦDS,in because EFSpump was a significant contributor to EDS,
net
in , as shown in Eq. (18).
Fig. 8 shows the calculated water flux distribution inside the module
for different inlet flow rates. In this figure, the change in the water flux
was shown using different colors. When the inlet flow rates were quite
low, EDS, innet for modules (1) and (2) were > 2.0 MJ·m−3, as shown in
139
Desalination 447 (2018) 133–146
the EDS, innet values inside the figures. However, in this case, the po-
tential module performance was not fully realized because much of the
HF membrane did not allow water permeation, as shown in the blue-
colored regions where no water flux was observed. When the DS-side
flow rate increased, the working membrane area for water permeation
expanded significantly along the r-axis, but EDS, innet decreased because
the FS-side flow rate was low compared with the water flux. In addition,
when the FS-side flow rate increased, EDS, innet in both modules reached
a certain value and then decreased. The optimum values for QFS,in at
QDS,in = 8 L·min−1 were approximately 5 and 9 L·min−1 for modules
(1) and (2), respectively. This difference was mainly due to the differ-
ence in the water permeability of the HF modules. On the other hand,
further increases in QFS,in led to decreasing EDS, innet, due to the in-
creased pressure drop. For all calculation conditions, areas of in-
effective water permeation (low r-axis and high z-axis areas, at the
lower right sides of the graphs in Fig. 8) existed in the HF modules,
meaning that the current HF modules would need to be improved by
modifying the module dimensions to achieve higher efficiency and
performance.
4.4. Optimum dimensions of the HF modules
To improve the performance of the HF modules by decreasing the
unused membrane area indicated by the blue-colored regions in Fig. 8
(areas of no water flux within the modules), simulation-based proposals
for the optimum dimensions of the HF modules were tested. HF mod-
ules with different module dimensions were proposed virtually to es-
timate the effects of the module dimensions on the PRO performance
Y. Tanaka et al. Desalination 447 (2018) 133–146

Fig. 6. EDS, in
net
of module (1) and (2) as a function of PDS,in for different inlet flow rates.

and efficiency. To mitigate the no-flux areas at the lower r-axis and same as the original module (module (a)). This module was abbreviated
higher z-axis areas, the z-axis length was first shortened by half. This as module (c). We then estimated the performance improvement of
module was abbreviated as module (b). An additional modification modules (b) and (c), as compared to the original, module (a). Modules
increased the r-axis dimension while keeping the module volume the (a), (b), and (c) each had two types of HFs. Therefore, data from six

140
Y. Tanaka et al. Desalination 447 (2018) 133–146

Fig. 7. Effects of the inlet flow rates (QFS,in and QDS,in) and the inlet flow ratio of DS (ΦDS,in) on EDS, innet under constant PDS,in values of 20 and 10 bar for modules (1)
and (2), respectively. The units of QDS,in are L·min−1. The theoretical maximum, EDS, intheo., is also shown as a dotted line.

Fig. 8. Distribution of water flux within (a) module (1) and (b) module (2) under different inlet flow rate conditions. (For interpretation of the references to color in
this figure, the reader is referred to the web version of this article.)

141
Y. Tanaka et al. Desalination 447 (2018) 133–146

Table 3
Specifications of the HF membrane modules used in the simulation.
Module length Module radius Number of fibers Effective membrane area Packing density Wound number Note
[m] [m] [−] [m2] [%] [−]

Module (1)(a) 0.58 0.117 220,000 70.5 54.2 2 Original

Module (1)(b) 0.29 0.117 220,000 38.4 54.2 1 Proposed
Module (1)(c) 0.29 0.165 434,000 87.2 54.2 1 Proposed
Module (2)(a) 0.58 0.117 165,600 65.6 43.2 2 Original
Module (2)(b) 0.29 0.117 165,600 32.1 43.2 1 Proposed
Module (2)(c) 0.29 0.165 333,000 73.1 43.2 1 Proposed

net
Fig. 9. Effect of module dimensions on the EDS, in .

modules are shown in Table 3.
Fig. 9 shows the simulated performance of the respective modules
with different ΦDS,in for the cases of DS flow rates of 4 or 8 L·min−1. The
applied hydraulic pressure differences for modules (1) and (2) were set
as 20 and 10 bar, respectively. In the case of module (b), because the
effective membrane area became approximately half of that of module
(a), due to the half-length of module (a), the DS flow rate was set to be
half (2 or 4 L·min−1) of those for modules (a) and (c), for a fair com-
parison (the DS flow rate per effective membrane area was adjusted to
be the same). The theoretical maximum value was also calculated by
assuming a perfect PRO membrane for comparison. Fig. 9 indicated that
EDS, innet increased due to shortening the module length by half (module
(b)), and slightly increased due to the increased r-axis dimension of
module (c), as compared with the original, module (a). The former and
latter improvements were due to the mitigation of the FS-side pressure
drop and sufficient water permeation, respectively.
Fig. 10 shows the water flux distributions in the different types of
modules based on module (1) (modules (1)(a), (1)(b), and (1)(c)). The
DS flow rates for modules (1)(a) and (1)(c) were both adjusted to
8 L·min−1 and the FS flow rates were set to 5 and 8 L·min−1 for modules
(a) and (c), respectively, where the values of φDS,in were 0.62 and 0.5,
respectively. As mentioned above, the DS and FS flow rates for module
(b) were set to half that of module (a) (4 L·min−1 and 2.5 L·min−1,
respectively). In the case of module (1)(a), ineffective membrane area
(blue-colored region) partially existed when EDS, innet reached its max-
imum. Although this ineffective area could be decreased by increasing
the FS inlet flow rate to 8 L·min−1, further increase of the FS inlet flow
would also lead to an unfavorable increase in the FS side pressure drop
and subsequent decrease in EDS, innet. Therefore, this result indicates that
an ineffective area clearly exists, especially at the high z-axis of the
original HF module. On the other hand, in the case of modules (1)(b)
and (1)(c), the portions of ineffective membrane areas could be de-
creased by shortening along the z-axis and increasing the r-axis di-
mension of the module, as shown in Fig. 10(b) and (c). In addition,
enhancement of EDS, innet can be achieved, as shown in Fig. 10(c), when
increasing the FS inlet flow rate (QFS,in = 8 L·min−1) by optimization of

142
Y. Tanaka et al. Desalination 447 (2018) 133–146

Fig. 10. Distribution of water flux within the modules under different inlet flow rate conditions for (a) module (1)(a), (b) module (1)(b), and (c) module (1)(c). (For
interpretation of the references to color in this figure, the reader is referred to the web version of this article.)

the module dimensions, given the same module volume. Therefore, this
simulation indicated that the shorter module with a larger diameter was
more suitable from the view point of both higher EDS, innet and more
effective use of all the membrane area within the module. The esti-
mated EDS, innet of module (1)(c) (1.06 W·m−3) was approximately 30%
higher than that of module (1)(a) (0.84 W·m−3) at QDS,in = 8 L·min−1
and QFS,in = 8 L·min−1.
Fig. 11 also shows the distribution of the water flux for the different
module types based on module (2) (modules (2)(a), (2)(b), and (2)(c)).
In these modules, because the water permeability was greater than in
module (1), shortening in the z-axis direction and widening in the r-axis
direction did not alter the ineffective membrane area substantially, as
shown in Fig. 11(b) and (c). However, the simulation results indicated
that the shorter and larger-diameter module also increased EDS, innet, as
compared with the original. In this case, the migration of the FS side
pressure drop was the main reason for achieving the increase in the EDS,
net
in . The pressure drops in all the modules are listed in Table 4. The
pressure drops of modules (2)(b) and (2)(c) were 1.2 bar and were less
than half of that of module (2)(a). The estimated EDS, innet of module (2)
(c) (1.02 W·m−3) was approximately 43% greater than that of module
(2)(a) (0.71 W·m−3) at QDS,in = 8 L·min−1 and QFS,in = L·min−1. Not-
withstanding the higher water permeability of module (2) as compared
to module (1), the estimated EDS, innet of module (2)(c) was slightly
lower than that of module (1)(c), due to the lower hydraulic pressure
(10 bar) than in the module (1) case (20 bar) necessitated by the lower
hydraulic pressure tolerance of the HFs, as mentioned previously.

143
Y. Tanaka et al. Desalination 447 (2018) 133–146

Fig. 11. Distribution of water flux within the modules under different inlet flow rate conditions for (a) module (2)(a), (b) module (2)(b), and (c) module (2)(c).

Table 4 membrane areas were decreased (for module (1), QDS,in = 8 L·min−1,
The pressure drops in all modules. and QFS,in = 8 L·min−1, and, for module (2), QDS,in = 8 L·min−1 and
(a) (b) (c)
QFS,in = 12 L·min−1), were approximately 9.3%, 11.7%, 7.8%, and
11.3%, respectively, of the energy consumption of current RO desali-
Module (1) 3.7 bar 1.4 bar 1.4 bar nation processes (9.0 MJ·m−3 (2.5 kWh·m−3)), and approximately 22%,
QDS,in = 8 L·min−1 QDS,in = 4 L·min−1 QDS,in = 8 L·min−1 28%, 19%, and 27%, respectively, of the theoretical energy consump-
QFS,in = 8 L·min−1 QFS,in = 4 L·min−1 QFS,in = 8 L·min−1
tion of RO with a recovery ratio of 50% (3.78 MJ·m−3
Module (2) 2.9 bar 1.2 bar 1.2 bar
QDS,in = 8 L·min−1 QDS,in = 4 L·min−1 QDS,in = 8 L·min−1 (1.03 kWh·m−3)).
QFS,in = 12 L·min−1 QFS,in = 6 L·min−1 QFS,in = 12 L·min−1

Although we ignored the mechanical efficiencies, such as of the
turbine, pump, and energy recovery device, in order to simplify module
performance estimation, the estimated EDS, innet for the optimum con-
ditions for modules (1)(a), (1)(c), (2)(a), and (2)(c), where the unused
5. Conclusions
The pilot scale module performance of two types of 5-inch HF
modules was experimentally and theoretically investigated under a
wide range of PRO operating conditions. For PRO performance esti-
mation, we modified the FCP model by combining PRO theory with

144
Y. Tanaka et al. Desalination 447 (2018) 133–146

water flux and salt leakage. The calculated results were in good bore inner bore-side of HF
agreement with the experimental data. shell outer shell-side of HF
Using the improved model, the estimation of the energy output in at inlet
performance of the 5-inch membrane modules revealed that 10.7 to out at outlet
15.7% (module (1)) and 9.2 to 14.2% (module (2)) of the energy could DS − FS difference between DS and FS
be recovered from the reverse osmosis seawater desalination process 0 initial
when the mechanical efficiency was ignored. Furthermore, in the ori-
ginal membrane modules, it was revealed that all the membrane area Acknowledgements
inside the module could not be effectively utilized, and water per-
meation was quite low in some areas inside the module. In order to This work was supported by a grant from the Ministry of Education,
overcome this problem, we proposed new types of modules with shorter Culture, Sports, Science and Technology (MEXT) for the Regional
lengths and larger diameters. These modules reduced the regions where Innovation Strategy Support Program of Japan and was partially sup-
the water was not sufficiently permeated and reduced the pressure ported by Grants-In-Aid from the Special Coordination Funds for
drop, which resulted in increased net energy output compared with the Promoting Science and Technology, Creation of Innovation Centers for
original modules. Thus, it was clarified that the structure of the mem- Advanced Interdisciplinary Research Areas (Innovative Bioproduction,
brane module, as well as the membrane performance, are quite im- Kobe), from the MEXT, Japan.
portant for enhancing the energy output from an osmotic process.
Appendix A. Supplementary data
Nomenclature
Supplementary data to this article can be found online at https://
A water permeability coefficient, L·m−2·h−1·bar−1 doi.org/10.1016/j.desal.2018.09.015.
Am effective membrane area, m2
B solute permeability coefficient, L·m−2·h−1 References
C concentration of solute, mol·L−1 (or mol·m−3)
Ddiff mutual diffusion coefficient of DS, m2·h−1·10−9 [1] C. Fritzmann, J. Löwenberg, T. Wintgens, T. Melin, State-of-the-art of reverse os-
(=μL·m−2·h−1) mosis desalination, Desalination 216 (2007) 1–76.
[2] R.K. McGovern, J.H. Lienhard V, On the potential of forward osmosis to en-
din inner diameter of hollow fibers, m ergetically outperform reverse osmosis desalination, J. Membr. Sci. 469 (2014)
dout outer diameter of hollow fibers, m 245–250.
EDS, innet net generated work per unit volume of DS, J·m−3-DS [3] B. Peñate, L. García-Rodríguez, Current trends and future prospects in the design of
seawater reverse osmosis desalination technology, Desalination 284 (2012) 1–8.
EDS, intheo. theoretical generated work per unit volume of DS, J·m−3-DS [4] K.H. Mistry, R.K. McGovern, G.P. Thiel, E.K. Summers, S.M. Zubair, J.H. Lienhard
Jw water flux, L·m−2·h−1 V, Entropy generation analysis of desalination technologies, Entropy 13 (2011)
Js solute flux, mol·m−2·h−1 1829–1864.
[5] M. Kurihara, H. Sakai, A. Tanioka, H. Tomioka, Role of pressure-retarded osmosis
k mass transport coefficient, L·m−2·h−1 (PRO) in the mega-ton water project, Desalin. Water Treat. 57 (2016)
P hydraulic pressure, bar 26518–26528.
Q volumetric flow rate, L·h−1 (or L·min−1) [6] M. Kurihara, M. Hanakawa, Mega-ton Water System: Japanese national research
R gas constant, bar·L·mol−1·K−1 and development project on seawater desalination and wastewater reclamation,
Desalination 308 (2013) 131–137.
r coordinate in the radial direction of the module, m [7] A. Tanioka, Preface to the special issue on “Pressure Retarded Osmosis in Megaton
Re Reynolds number, − Water System Project”, Desalination 389 (2016) 15–17.
S structure parameter of the support layer, μm [8] H. Sakai, T. Ueyama, M. Irie, K. Matsuyama, A. Tanioka, K. Saito, A. Kumano,
Energy recovery by PRO in sea water desalination plant, Desalination 389 (2016)
Sc Schmidt number, − 52–57.
Sh Sherwood number, − [9] K. Saito, M. Irie, S. Zaitsu, H. Sakai, H. Hayashi, A. Tanioka, Power generation with
T solution temperature, K salinity gradient by pressure retarded osmosis using concentrated brine from SWRO
system and treated sewage as pure water, Desalin. Water Treat. 41 (2012) 114–121.
u linear velocity of solution, m·s−1 [10] F.T. Khaled Touati, Joon Ha Kim, Oscar Andres Alvarez Silva, Pressure Retarded
V volume of the solution, m3 Osmosis: Renewable Energy Generation and Recovery, Academic Press, 2017.
W wound number of fibers around central core tube, − [11] A. Altaee, A. Sharif, Pressure retarded osmosis: advancement in the process appli-
cations for power generation and desalination, Desalination 356 (2015) 31–46.
Wp maximum extractive work, J [12] I. Alsvik, M.-B. Hägg, Pressure retarded osmosis and forward osmosis membranes:
z coordinate in the axial direction of the module, m materials and methods, Polymers 5 (2013) 303–327.
[13] 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.
Greek letters
[14] N.Y. Yip, M. Elimelech, Thermodynamic and energy efficiency analysis of power
generation from natural salinity gradients by pressure retarded osmosis, Environ.
α parameter in Eq. (9), − Sci. Technol. 46 (2012) 5230–5239.
[15] A. Achilli, J.L. Prante, N.T. Hancock, E.B. Maxwell, A.E. Childress, Experimental
β parameter in Eq. (9), −
results from RO-PRO: a next generation system for low-energy desalination,
γ parameter in Eq. (9), − Environ. Sci. Technol. 48 (2014) 6437–6443.
ε porosity of the support layer, − [16] G. Han, S. Zhang, X. Li, T.-S. Chung, Prog. in pressure retarded osmosis (PRO)
μ viscosity, Pa·s membranes for osmotic power generation, Prog. Polym. Sci. 51 (2015) 1–27.
[17] N.Y. Yip, A. Tiraferri, W.A. Phillip, J.D. Schiffman, L.A. Hoover, Y.C. Kim,
ν van't Hoff factor, − M. Elimelech, Thin-film composite pressure retarded osmosis membranes for sus-
π osmotic pressure, bar tainable power generation from salinity gradients, Environ. Sci. Technol. 45 (2011)
ρ density, kg·m−3 4360–4369.
[18] S. Chou, R. Wang, L. Shi, Q. She, C. Tang, A.G. Fane, Thin-film composite hollow
τ tortuosity of the support layer, − fiber membranes for pressure retarded osmosis (PRO) process with high power
Φ initial flow ratio, − density, J. Membr. Sci. 389 (2012) 25–33.
[19] G. Han, P. Wang, T.S. Chung, Highly robust thin-film composite pressure retarded
osmosis (PRO) hollow fiber membranes with high power densities for renewable
Subscripts salinity-gradient energy generation, Environ. Sci. Technol. 47 (2013) 8070–8077.
[20] S. Sarp, Z. Li, J. Saththasivam, Pressure Retarded Osmosis (PRO): past experiences,
DS draw solution side current developments, and future prospects, Desalination 389 (2016) 2–14.
[21] C.F. Wan, T.-S. Chung, Osmotic power generation by pressure retarded osmosis
FS feed solution side
using seawater brine as the draw solution and wastewater retentate as the feed, J.
b at bulk

145
Y. Tanaka et al.
Membr. Sci. 479 (2015) 148–158.
[22] G. Han, S. Zhang, X. Li, T.-S. Chung, High performance thin film composite pressure
retarded osmosis (PRO) membranes for renewable salinity-gradient energy gen-
eration, J. Membr. Sci. 440 (2013) 108–121.
[23] Y.C. Kim, M. Elimelech, Potential of osmotic power generation by pressure retarded
osmosis using seawater as feed solution: analysis and experiments, J. Membr. Sci.
429 (2013) 330–337.
[24] M. Higa, D. Shigefuji, M. Shibuya, S. Izumikawa, Y. Ikebe, M. Yasukawa, N. Endo,
A. Tanioka, Experimental study of a hollow fiber membrane module in pressure-
retarded osmosis: module performance comparison with volumetric-based power
outputs, Desalination 420 (2017) 45–53.
[25] Y. Xu, X. Peng, C.Y. Tang, Q.S. Fu, S. Nie, Effect of draw solution concentration and
operating conditions on forward osmosis and pressure retarded osmosis perfor-
mance in a spiral wound module, J. Membr. Sci. 348 (2010) 298–309.
[26] Y.C. Kim, Y. Kim, D. Oh, K.H. Lee, Experimental investigation of a spiral-wound
pressure-retarded osmosis membrane module for osmotic power generation,
Environ. Sci. Technol. 47 (2013) 2966–2973.
[27] E. Desormeaux, Blue Tech Webinar Series: A Focus on Pressure Retarded Osmosis,
(March 20, 2014).
[28] M. Shibuya, M. Yasukawa, S. Goda, H. Sakurai, T. Takahashi, M. Higa,
H. Matsuyama, Experimental and theoretical study of a forward osmosis hollow
fiber membrane. Module with a cross-wound configuration, J. Membr. Sci. 504
(2016) 10–19.
[29] M. Sekino, Precise analytical model of hollow fiber reverse osmosis modules, J.
Membr. Sci. 85 (1993) 241–252.
[30] S. Kimura, S. Sourirajan, Analysis of data in reverse osmosis with porous cellulose
146
Desalination 447 (2018) 133–146
acetate membranes used, AICHE J. 13 (1967) 497–503.
[31] Q. She, X. Jin, C.Y. Tang, Osmotic power production from salinity gradient resource
by pressure retarded osmosis: effects of operating conditions and reverse solute
diffusion, J. Membr. Sci. 401-402 (2012) 262–273.
[32] M. Sekino, Study of an analytical model for hollow fiber reverse osmosis module
systems, Desalination 100 (1995) 85–97.
[33] A. Kumano, H. Matsuyama, Analysis of hollow fiber reverse osmosis membrane
module of axial flow type, J. Appl. Polym. Sci. 123 (2012) 463–471.
[34] M. Yasukawa, S. Mishima, Y. Tanaka, T. Takahashi, H. Matsuyama, Thin-film
composite forward osmosis membrane with high water flux and high pressure re-
sistance using a thicker void-free polyketone porous support, Desalination 402
(2017) 1–9.
[35] M. Shibuya, M. Yasukawa, T. Takahashi, T. Miyoshi, M. Higa, H. Matsuyama,
Effects of operating conditions and membrane structures on the performance of
hollow fiber forward osmosis membranes in pressure assisted osmosis, Desalination
365 (2015) 381–388.
[36] M. Shibuya, M. Yasukawa, T. Takahashi, T. Miyoshi, M. Higa, H. Matsuyama, Effect
of operating conditions on osmotic-driven membrane performances of cellulose
triacetate forward osmosis hollow fiber membrane, Desalination 362 (2015) 34–42.
[37] F. Helfer, C. Lemckert, Y.G. Anissimov, Osmotic power with pressure retarded os-
mosis: theory, performance and trends – a review, J. Membr. Sci. 453 (2014)
337–358.
[38] A. Achilli, T.Y. Cath, A.E. Childress, Power generation with pressure retarded os-
mosis: an experimental and theoretical investigation, J. Membr. Sci. 343 (2009)
42–52.