Article

[Link]/IECR

A Unified Model-Based Analysis and Optimization of Specific Energy

Consumption in BWRO and SWRO
Mingheng Li*
Department of Chemical and Materials Engineering, California State Polytechnic University, Pomona, California 91768, United States

ABSTRACT: A unified mathematical model is developed to describe both brackish water (BW) and seawater (SW) reverse
osmosis (RO) desalination processes. The model contains two dimensionless parameters: κ, quantifying the effect of retentate
pressure drop, and γ, reflecting the membrane capacity demand ratio. A γ vs. κ map derived from data obtained from RO
desalination plants worldwide is used to show that BWRO and SWRO are in completely different regions. The minimization of
specific energy consumption (SEC) is formulated and solved as a nonlinear optimization problem using representative values of γ
and κ in both BWRO and SWRO. Several computational case studies are carried out to demonstrate the differences in operating
BWRO and SWRO. Discussions are made on the effect of fractional recovery, RO configurations (single-stage, two-stage, and
two-stage with booster pump) as well as brine recirculation on SEC.

■ INTRODUCTION
Specific energy consumption (SEC), or pump energy
■ DIMENSIONLESS RO MODEL
On the basis of assumptions of Darcy’s law for mass transfer
consumption divided by water production rate, is a very and quadratic dependence of retentate pressure drop on
important topic in RO desalination.1,2 There have been velocity, the following mathematical model was derived to
significant research and development efforts in model-based describe a single-stage RO process14
analysis and optimization of RO design and operation.3−9
Model-based control has also been developed recently to dQ (x) ⎛ Q ⎞
− = ALp⎜ΔP − 0 Δπ0⎟
reduce SEC in a pilot-scale RO system10,11 dx ⎝ Q ⎠
In two previous papers, 12,13 the author provided a
comprehensive analysis of single- and multistage SWRO d(ΔP(x))
=−kQ 2
with/without energy recovery device (ERD) from first- dx
principles, based on the assumption of negligible retentate
Q (x ) =Q 0@x = 0
pressure drop. Several dimensionless parameters (γ, α, and Y)
were introduced to show the coupled behavior between ΔP(x) = ΔP0@x = 0 (1)
membrane property (area and permeability), feed conditions
(flow rate and salinity), and operating conditions (hydraulic where Q is the retentate flow rate, −dQ is the flow rate of water
pressure difference and fractional recovery). Later on, the across membrane of area dA (dA = Adx; x is the distance from
retentate pressure drop effect was explicitly accounted for in the entrance normalized by the pressure vessel length, and A is the
optimization model for both single-train and multitrain BWRO total area of all RO elements in this stage), Lp is the membrane
processes.14,15 A validation of model-based optimization results hydraulic permeability, ΔP and Δπ are the differences in
was carried out in a BWRO plant in Chino, California, and a hydraulic pressures and osmotic pressures across the
10% decrease in SEC and a substantial reduction in brine membrane, respectively. Δπ0, Q0, and ΔP0 are Δπ, Q, and
disposal cost were demonstrated.16 ΔP at the stage entrance. k is a coefficient describing the
As a continuation of the author’s previous work, this paper pressure drop along the pressure vessel in the retentate stream.
aims to utilize a unified dimensionless model to explain the Note that the pressure differential in the permeate stream is
similarities and differences between SWRO and BWRO. The typically small. More detailed assumptions and justifications of
model is based on several dimensionless parameters, including eq 1 are discussed in the author’s previous work.14 This coupled
κ, quantifying the effect of retentate pressure drop, and γ, a differential equation captures the profile of driving force and
previously derived parameter reflecting the membrane capacity water flux, and is more advanced than standard membrane
demand ratio.12 Data from both BWRO and SWRO plants projection model in literature, which uses averaged flux and
worldwide are used to show the typical ranges of γ and κ. Using driving force.
representative values of γ and κ in both BWRO and SWRO, a Defining α = Δπ0/ΔP0, γ = ALpΔπ0/Q0, κ = kQ20/Δπ0, q =
mathematical model is formulated to reduce SEC by optimizing Q/Q0, p = ΔP/ΔP0, the following equation is derived:
a dimensionless pressure parameter α. The optimal SECs in
both BWRO and SWRO using three different RO config- Received: September 18, 2013
urations (single-stage, two-stage, and two-stage with booster Revised: November 8, 2013
pump) are compared. The effect of brine recirculation on SEC Accepted: November 15, 2013
is also discussed. Published: November 18, 2013

© 2013 American Chemical Society 17241 [Link]/10.1021/ie4031083 | Ind. Eng. Chem. Res. 2013, 52, 17241−17248
Industrial & Engineering Chemistry Research Article

dp(x) relationships between p and q at the stage level and those at the
= −καq2(x) system level are shown as follows
dx
dq(x) ⎛ p(x ) 1 ⎞ i−1
= −γ ⎜ − ⎟ p(x) = pi (x) ∏ pk (k), i − 1 < x ≤ i , i = 1, 2, ..., N
dx ⎝ α q(x) ⎠
k=0
p(x ) = 1, @x = 0 i−1
q(x) = qi(x) ∏ qk (k), i − 1 < x ≤ i , i = 1, 2, ..., N
q(x) = 1, @x = 0 (2) k=0
(5)
An integration of eq 2 would provide the profiles of p(x) and
q(x) along the RO stage. The fractional recovery of water is where p0(0) ≡ q0(0) ≡ 1. The system recovery at the end of all
then calculated as Y = 1 − q(1), where q(1) is q evaluated at x = stages Y = 1 − Πi N= 1qi(i).
1, or the end of this stage. The parameters αi, γi, and κi should be updated as eq 4 is
It is seen from eq 2 that the RO system is characterized by solved sequentially from stage to stage. For example, in a typical
three different dimensionless parameters α, γ, and κ, all two-stage RO process with a 2:1 array to achieve a high system
described in the form of ratios. As discussed in previous system recovery and to maintain a relatively uniform feed flow
publications,12,13 α is the ratio of osmotic pressure difference to in each stage,14,17 A2 = A1/2, Ac2 = Ac1/2, k2 = 22k1, Δπ1 = Δπ0/
hydraulic pressure difference across the membrane at the q1(1), ΔP1 = ΔP0/p1(1), Q1 = Q0q1(1). As a result, γ2 = γ1/
entrance of the stage, and γ is the ratio of flow rate across the 2q21(1), α2 = α1/q1(1)/p1(1), κ2 = 4q31(1)κ1. Once the first stage
membrane assuming a constant drive force of Δπ0 to the is solved, α2, γ2, and κ2 are determined so that the second stage
retentate flow rate at the entrance of the stage. γ may be can be solved in a similar way. The system recovery at the end
considered as a measure of the membrane capacity demand of the second stage is calculated as Y = 1−q1(1)q2(2).
ratio. κ introduced in this work is the ratio of maximum The SEC normalized by the feed osmotic pressure, or NSEC,
retentate pressure drop (i.e., the theoretical pressure drop when is a dimensionless number12
there is no water recovery, or the retentate flow rate is always
the same as the intake flow rate) to the osmotic pressure Q 0ΔPpump
difference at the entrance of the stage. It is a measure of driving NSEC =
force loss that adversely affects the water production. According ηpump(Q 0Y )Δπ0 (6)
to fluid mechanics of channel flow, κ can be described as
follows where ΔPpump is the pressure increase across the pump, ηpump is
the pump efficiency. Note that ΔP0 = Pbp + ΔPpump − Pp, where
κ = kQ 02/Δπ0 = 2f (Lpv /DH)ρ(Q 0/Ac)2 /Δπ0 (3) Pbp is the feed pressure before pump, and Pp is the permeate
pressure. Therefore
where f is the Fanning friction coefficient, DH is the hydraulic
diameter between membrane sheets, Ac is the total cross- 1/α − (Pbp − Pp)/Δπ0
sectional area in the RO stage for retentate flow, and Lpv is the NSEC= N
total length of the pressure vessel that houses the membrane ηpump(1 − ∏i = 1 qi(i)) (7)
elements in series. Apparently, κ will reduce if more pressure
vessels are put in parallel in a RO stage, or fewer membrane The minimization of NSEC in a single- or multistage RO is
elements are enclosed in a pressure vessel. then formulated as an optimization problem as follows
The introduction of dimensionless parameters simplifies the
analysis of RO process as the results are scalable. Because 1/α1 − (Pbp − Pp)/Δπ0
BWRO and SWRO differ in Δπ0 (around 10−20 psi in BWRO min J = N
α1 ηpump(1 − ∏i = 1 qi(i))
and 390 psi in SWRO), they may exhibit different process
characteristics and require different design and operation s. t .
strategies.
Equation 2 can be easily extended to describe RO networks dpi (x)
= −κiαiqi2(x), i = 1, 2, ..., N
with complicated configurations. For example, a RO system dx
with N stages in series may be modeled as follows
dqi(x) ⎛ p (x ) 1 ⎞⎟
= −γi⎜⎜ i − , i = 1, 2, ..., N
dpi (x)
= −κiαiqi2(x), i = 1, 2, ..., N dx ⎝ αi qi(x) ⎟⎠
dx
pi (x) = 1, @x = i − 1, i = 1, 2, ..., N
dqi(x) ⎛ p (x ) 1 ⎞⎟
= −γi⎜⎜ i − , i = 1, 2, ..., N
qi(x) ⎟⎠
qi(x) = 1, @x = i − 1, i = 1, 2, ..., N (8)
dx ⎝ αi
pi (x) = 1, @x = i − 1, i = 1, 2, ..., N The optimization of a RO network with more complicated
configurations can be formulated using methods developed in
qi(x) = 1, @x = i − 1, i = 1, 2, ..., N (4) the author’s previous work.13 For a two-stage RO with
interstage booster pump, the optimization problem (neglecting
where pi and qi are defined based on the ratio of ΔP(x) and pump efficiency and assuming pbp = pp for simplicity) may be
Q(x) in RO stage i to the ones at the entrance of stage i. The formulated as follows
17242 [Link]/10.1021/ie4031083 | Ind. Eng. Chem. Res. 2013, 52, 17241−17248
Industrial & Engineering Chemistry Research Article

1/α1 + 1/α2 − p1 (1)q1(1) in parentheses. In all calculations, osmotic pressure is estimated

min J = based on TDS data using an online calculator ([Link]
α1, α2 1 − q1(1)q2(2) [Link]/calculators/osmotic/[Link]), if it
s. t . is not provided in the literature.
A significant difference between BWRO and SWRO is
dpi (x) observed. For BWRO, γ1 is typically about 0.1 (even γtotal is less
= −κiαiqi2(x), i = 1, 2
dx than 0.2 if multiple stages are used), whereas κ1 is between 1
and 6. For SWRO, γ1 is between 0.5 and 1.5, whereas κ1 is less
dqi(x) ⎛ p (x ) 1 ⎞⎟ than 0.2. The differences are presented in the γ vs. κ map in
= −γi⎜⎜ i − , i = 1, 2
dx ⎝ αi qi(x) ⎟⎠ Figure 1. These differences are mainly caused by the fact that

pi (x) = 1, @x = i − 1, i = 1, 2
qi(x) = 1, @x = i − 1, i = 1, 2
γ1 = (2/3)γtotal

γ2 = γ1/2/q12(1)

κ2 = 4κ1q13(1)
α2 ≤ α1/p1 (1)/q1(1)
−αi ≤ 0, i = 1, 2
αi ≤ 1, i = 1, 2 (9)
In eq 9, γtotal is defined based on the total area of all RO
elements in both stages (γtotal = AtotalLpΔπ0/Q0). Note that
while Atotal = A1 + A2, γtotal ≠ γ1 + γ2. Inequality constraint α2 ≤ Figure 1. γ and κ in BWRO and SWRO.
α1/p1(1)/q1(1) is included to ensure non-negative pressure
increase across the interstage pump.
The optimization models described by eqs 8 and 9 can be the osmotic pressure of seawater is at least 1 order of
solved using common constrained multivariable nonlinear magnitude higher than the one of brackish water. Current
optimization packages, e.g., fmincon in Matlab. BWRO plants are not built with a large γ because of constraints

■ RESULTS AND DISCUSSION

Analysis of BWRO and SWRO Plant Data. Data of
in capital cost. For example, the WMZM plant in Jordan has
three BWRO stages with a total of 791 elements, but γtotal is
estimated to be 0.12 base on the operational data in one of the
BWRO and SWRO plants worldwide14,18−29 are analyzed and skids.27 The differences in γ and κ imply different process
the parameters α, γ, and κ are either calculated directly based characteristics as well as design and operation strategies in
on their definitions or correlated to match plant measurements SWRO and BRWO. It is worth noting that in a two-pass
including pressure differentials and recoveries. The results are seawater treatment plant, the first pass is SWRO, whereas the
summarized in Table 1. γ, κ, and α reported here are based on second-pass is BWRO.30−32
the first stage for plants employing multistage operation as they Optimization of NSEC in Single-Stage BWRO and
change from stage to stage. For these plants, γtotal are provided SWRO. The optimization problem described by eq 8 is solved
for single-stage RO with representative values of γ and κ in
Table 1. Correlation of Plant Data with Mathematical Model BWRO and SWRO. For demonstration purposes, it is assumed
that Pp = Pbp. Moreover, the effect of pump efficiency is
plant RO type γ1 (γtotal) κ1 α1 Y neglected. Interested readers are referred to the author’s
Carlsbad Pilot, USA 18
SWRO 1.43 0.13 0.49 0.50
previous work14 for optimization accounting for pump
Tuas, Singapore19 SWRO 0.66 0.13 0.46 0.45
characteristic maps and other process details.
Tuas, Singapore20 SWRO 0.86 0.12 0.50 0.45
The optimal NSEC and the corresponding α and Y are
Jeddah, Saudi SWRO 0.86 0.08 0.53 0.41
shown in Figures 2, 3, and 4. It is seen that the optimal NSEC
Arabia21,22 decreases as κ decreases and γ increases. When both κ and 1/γ
North Obhor, Saudi SWRO 0.88 0.06 0.57 0.36 go to zero, NSEC reaches 4 and Y = 50% by applying ΔP0 =
Arabia23 2Δπ0. This is well-known in the RO literature.12 However, in
Perth, Australia24 SWRO 0.61 0.05 0.46 0.45 real-world RO plants, γ is always finite (and may be very small
ADC Pilot, USA25 SWRO 0.56 0.04 0.45 0.45 in BWRO) because of considerations in capital cost. Even
WMZM, Jordan (three- BWRO 0.07 (0.12) 4.15 0.09 0.88 though κ is small in SWRO, it is substantial in BWRO. A larger
stage)26,27
γ allows the system to be operated closer to the thermodynamic
Chino, USA (two- BWRO 0.04 (0.05) 5.44 0.05 0.81
stage)14 limit, leading to a lower NSEC.12,13 A larger κ, however,
James City (two-stage), BWRO 0.10 (0.15) 1.24 0.13 0.78 reduces the driving force for permeate flow, leading to a higher
USA28 NSEC. When γ decreases (less membrane capacity) or κ
Ocean Reef Club, BWRO 0.08 1.06 0.10 0.65 increases (more pressure drop), a higher recovery and a greater
USA29 applied pressure are suggested in order to achieve the NSEC at
17243 [Link]/10.1021/ie4031083 | Ind. Eng. Chem. Res. 2013, 52, 17241−17248
Industrial & Engineering Chemistry Research Article

its lowest attainable level. It is worth noting that the maximum

recovery is limited by the ability to produce potable quality
permeate in SWRO33 and the occurrence of silica precipitation
in BWRO.34
From Figures 2, 3, and 4, one can read optimal operating
conditions for typical BWRO and SWRO processes using
representative values of γ and κ. The results are shown in Table
2. Apparently, NSEC is only slightly above 4 in SWRO but
much larger in BWRO. To achieve the optimal NSECs, ΔP0 is
2.3 and 17 times Δπ0 in SWRO and in BWRO, leading to a
significant difference in water recoveries. The optimal recovery
in BWRO is much larger. The dimensionless driving forces at
the entrance and exit of the stage (p/α − 1/q) are provided in
the same table. It appears that SWRO is operated near the
thermodynamic limit, but BWRO is far away from it.
Comparison of SEC between Single-Stage and Two-
Figure 2. Optimal NSEC as a function of κ and γ in single-stage RO Stage Designs. Both single-stage and two-stage designs have
without ERD. been used in RO operation, and there are still debates on which
one is more efficient from a viewpoint of energy con-
sumption.35 In this subsection, several case studies will be
used to clarify the effect of RO configurations on SEC in both
BWRO and SWRO.
For BWRO, the effect of retentate pressure drop on water
recovery is first quantified before the detailed comparisons.
Equation 2 was applied to an industrial two-stage BWRO train
in Chino, California.14 This BWRO train has 28 pressure
vessels in parallel in the first stage and 14 in the second stage.
Each pressure vessel houses 7 Dow FILMTEC BW30−400 RO
elements in series. The BWRO train does not have an
interstage booster pump or ERD. It was derived from plant data
that k1 = 2.12 × 10−5 psi/gpm2 and Lp is 0.11 gfd/psi.14 The
model predictions using these two parameters were able to
match plant data collected in a wide range of operating
conditions.16 On the basis of these parameters, it is calculated
Figure 3. Optimal applied pressure as a function of κ and γ in single-
that κ1 = 5.5 and γ1 = 0.035.
stage RO without ERD. A comparison of simulation results with/without retentate
pressure drop is shown in Figure 5. When the pressure drop
effect is completely ignored (or assuming κ1 = κ2 = 0), the
calculated recovery is 91%, leading to an over prediction of
about 12% compared to its measured value of 81%. This case
study indicates that the retentate pressure drop should not be
ignored in BWRO where κ is large.
With the above information, one would expect that if all 42
pressure vessels are laid out in parallel in just one stage, this
BWRO plant could be more energy-efficient because of
reduced retentate pressure drop. The detailed comparison is
shown in Figures 6 and 7. The overall retentate pressure drop
ratio is only 4% in the single-stage design as compared to 24%
in the two-stage design. There are two factors that contribute to
such a significant difference. First, the channel area for retentate
flow in the single-stage configuration is 1.5 and 3 times of the
one in the first and second stage in the two-stage design,
respectively. The increased channel area leads to a reduced
Figure 4. Optimal recovery as a function of κ and γ in single-stage RO retentate velocity along the pressure vessel. Second, the total
without ERD. pressure vessel length in the single-stage design is only a half of

Table 2. Comparison between Single-Stage BWRO and SWRO under typical Operation Conditions

RO type γ κ αopt NESCopt Yopt p(0)/α−1/q(0) p(1)/α−1/q(1)

BWRO 0.05 1 0.059 22.9 0.75 16.1 12.7
SWRO 1 0.1 0.438 4.2 0.54 1.28 0.05
Thermodynamic Limit12 ∞ 0 0.5 4 0.5 1 0

17244 [Link]/10.1021/ie4031083 | Ind. Eng. Chem. Res. 2013, 52, 17241−17248

Industrial & Engineering Chemistry Research Article

calculation shows that a ΔP0 of 163 psi (10% lower than 180
psi in the current two-stage design) could keep the 81%
recovery with the same Q0. Therefore, the single-stage
configuration is more effective than its two-stage counterpart
in BWRO from a viewpoint of SEC (about 10% lower in this
case). However, it should be noted that the two-stage design
has a less drastic change in retentate flow per stage and less
tendency for concentration polarization36 because of a higher
velocity.
The same conclusion may be drawn for SWRO, even though
the difference between single-stage and two-stage configura-
tions is small, given the fact that κ is close to zero in SWRO.
However, it is worth pointing out that when an interstage
booster pump is used in conjunction with the two-stage design
in SWRO, it would be advantageous over the single-stage
design, according to the author’s published theoretical
Figure 5. Effect of retentate pressure drop on water recovery in an analysis.12,13 The computational analysis is based on the
industrial two-stage BWRO train without interstage booster pump. assumption κ = 0. The works also show that when γtotal < 0.2,
the optimal solution is corresponding to a zero pressure
increase in the interstage booster pump, or the two-stage
configuration is essentially one-stage. There are industrial cases
to substantialize reduced SEC in SWRO brine conversion using
interstage booster pumps.37 For example, in the SWRO plant in
Las Palmas III, Spain, an increase of 8% in the recovery of and
an improvement of 12.65% in the specific energy consumption
were observed after interstage booster pumps were installed.38
To theoretically validate whether a booster pump would
always reduce SEC for a nonzero κ in SWRO and BWRO, the
optimization problems described by eqs 8 and 9 are solved for
three different configurations (single-stage, two-stage, and two-
stage with booster pump, see Figure 8) using representative

Figure 6. Comparison of dimensionless applied pressure and retentate

flow between single-stage and two-stage designs in BWRO.

Figure 8. Schematic of three different RO configurations used in

industry.

values of γ and κ. In order to have a fair comparison, all three

configurations have exactly the same membrane properties and
flow characteristics. These mean that γtotal = AtotalLpΔπ/Q0 is
the same in all configurations. However, it should be noted that
if the pressure loss parameter is κ in the first stage of a two-
stage RO, it becomes (2/3)2κ in a single-stage RO because of
Figure 7. Comparison of dimensionless driving force between single- increased cross-sectional channel area for retentate flow (based
stage and two-stage designs in BWRO. on eq 3).
The optimized NSECs and corresponding recoveries for
the one in the two-stage design. A less retentate pressure drop BWRO and SWRO are shown in Figures 9 and 10. For BWRO,
results in a higher driving force for water production (see the optimal NSECs with/without a booster pump are almost
Figure 7). It is shown that the predicted water recovery in the the same. Both two-stage designs have a higher NSEC than the
single-stage design is 89% with the same Q0 and ΔPpump. To one in single-stage design at their optimal conditions.
maintain the same production, Q0 and/or ΔPpump may be Therefore, installing a booster pump would not help reduce
lowered in the single-stage configuration. For example, the SEC in BWRO. In SWRO, the two-stage with booster pump
17245 [Link]/10.1021/ie4031083 | Ind. Eng. Chem. Res. 2013, 52, 17241−17248
Industrial & Engineering Chemistry Research Article

Figure 9. Optimal (a) NSECs and (b) recoveries in BWRO for single- Figure 10. Optimal (a) NSECs and (b) recoveries in SWRO for
stage, two-stage, and two-stage with booster pump using κ = 1. single-stage, two-stage, and two-stage with booster pump using κ = 0.1.

Figure 11 shows an example in which an ERD (PX by Energy

design outperforms the other two designs at a large γtotal (e.g., Recovery Inc.) is intentionally unbalanced in a BWRO system
0.8 or greater in this case). However, the single-stage may be
the most energy efficient at a γtotal smaller than 0.8. Calculation
also shows that this critical γtotal reduces from 0.8 to 0.7 when κ
decreases from 0.1 to 0.05. In both BWRO and SWRO studied
here, the two-stage RO without booster pump is always no
better than the other two configurations from a viewpoint of
SEC.
The above comparisons among three RO configurations are
made at their best operating conditions. It is possible that one
design at its optimal condition is more energy-effective than
another at a nonoptimal condition. Moreover, because the two-
stage designs (with/without booster pumps) have higher
recoveries at its optimal condition than the one in the single-
stage, adding a constraint of maximum recovery may also
change the comparison results.
Effect of Brine Recirculation on SEC in BWRO. The use Figure 11. Diagram of RO with ERD.
of ERD is a proven technology to reduce SEC in SWRO.20,21,39
It has also been recently demonstrated in BWRO.29 However, which yields an overall system recovery (F divided by A) of
in some cases there is leakage of brine into the feed in the ERD, 85% and 2000 TDS feedwater, whereas the membrane recovery
which may increase the feed salinity and reduce the driving (F divided by E) is at 65% and 4886 TDS feedwater.40
force for separation. The same issue exists in brine recirculation The pressures and flows in all locations are shown in Table 3.
systems where energy contained in the brine is recovered. In Based on these data, it can be derived that α = 0.232, γ = 0.268,
this subsection, an example is used to quantify the effect of and κ = 0.426. If RO properties are the same when ERD is
brine feed mixing on SEC in BWRO. used, α = 0.092, γ = 0.138, and κ = 0.638 because γ ∝ Δπ0/Q0
17246 [Link]/10.1021/ie4031083 | Ind. Eng. Chem. Res. 2013, 52, 17241−17248
Industrial & Engineering Chemistry Research Article

Table 3. Comparison of Single-Stage BWRO with/without ERD

A B C D E F G H
with PX40
flow (gpm) 1774 250 1524 775 2299 1500 799 274
pressure (psi) 25 25 241 221 241 5 231 10
TDS (mg/L) 2000 2000 2000 10563 4886 92 14000 14000
without PX
flow (gpm) 1774 1774 1774 1604 170 170
pressure (psi) 25 241 241 5 237 237
TDS (mg/L) 2000 2000 2000 92 20000 20000

and κ∝Q20/Δπ0. The model-predicted pressures and flow rates retentate pressure drop. However, with an interstage
based on the new values of α, γ, and κ are shown in Table 3 for booster pump, two-stage may be more energy efficient
a comparison. than single-stage in SWRO. In both BWRO and SWRO
Without ERD, the recovery predicted by the model is 90% studied in this paper, the two-stage RO without booster
(see Figure 12), implying that the pump power may be reduced pump is always no better than the other two
configurations from a viewpoint of SEC.
• In brine recirculation and/or ERD, mixing of brine and
feed may increase salinity, adversely affecting SEC. A
trade-off between increased feed salinity and increased
energy recovery should be carefully evaluated.

■ AUTHOR INFORMATION
Corresponding Author
*E-mail: minghengli@[Link]. Phone: +1-909-869-
3668. Fax: +1-909-869-6920.
Notes
The authors declare no competing financial interest.

■
α
NOMENCLATURE
Δπ0/ΔP, dimensionless
Figure 12. Effect of brine leakage in ERD on retentate pressure drop Δπ0 osmotic pressure difference across the membrane at
and recovery. entrance, dimensionless
γ ALpΔπ0/Q0, dimensionless
to maintain the same production. Without blending the fresh κ κQ20/Δπ0, dimensionless
feed with the highly concentrated brine stream, a lower osmotic π osmotic pressure, bar
pressure is maintained in the retentate. Moreover, a smaller A membrane area, m2
feed flow rate (1774 gpm instead of 2299 gpm) leads to a Ac cross-sectional area for retentate flow, m2
smaller pressure drop in the retentate stream. Both factors DH hydraulic diameter between membrane sheets, m
contribute to a larger driving force for permeate production. f Fanning friction coefficient, dimensionless
This case indicates that a trade-off between reduced driving k pressure drop coefficient, bar/(m3/s)2
force and increased energy recovery should be carefully Lp hydraulic permeability of membrane, m/s/bar
evaluated from a viewpoint of SEC when brine recirculation Lpv length of pressure vessel, m
is used. This is especially important for BWRO because the P pressure, bar
pressure drop effect is significant. P ΔP/ΔP0, dimensionless

■ CONCLUSIONS
The following conclusions are drawn based on the computa-
Pb permeate pressure, bar
Pbp pressure before pump, bar
Q flow rate, m3 s−1
tional studies of RO desalination in this work: q Q/Q0, dimensionless
Y recovery, dimensionless

■
• Even though BWRO and SWRO may be described by
the same mathematical model, their process character-
REFERENCES
istics are different because they are in completely
different regions in the γ vs. κ map. The pressure drop (1) Semiat, R. Energy issues in desalination processes. Environ. Sci.
effect on RO performance cannot be ignored for BWRO. Technol. 2008, 42, 8193−8201.
(2) Elimelech, M.; Phillip, W. A. The future of seawater desalination:
• In current RO plants, SWRO may be operated near the Energy, technology, and the environment. Science 2011, 333, 712−717.
thermodynamic limit to reduce SEC. However, BWRO is (3) Zhu, A.; Christofides, P. D.; Cohen, Y. Effect of thermodynamic
operated far away from the thermodynamic limit because restriction on energy cost optimization of RO membrane water
of a small γ. desalination. Ind. Eng. Chem. Res. 2009, 48, 6010−6021.
• Without interstage booster pumps, single-stage is more (4) Oh, H.; Hwang, T.; Lee, S. A simplified simulation model of RO
energy efficient than two-stage in BWRO because of less systems for seawater desalination. Desalination 2009, 238, 128−139.

17247 [Link]/10.1021/ie4031083 | Ind. Eng. Chem. Res. 2013, 52, 17241−17248

Industrial & Engineering Chemistry Research Article

(5) Skiborowski, M.; Mhamdi, A.; Kraemer, K.; Marquardt, W. (28) 5MGD Ground Water Reverse Osmosis Treatment; James City
Model-based structural optimization of seawater desalination plants. Service Authority Five Forks Water Treatment Facility: James City
Desalination 2012, 292, 30−44. County, VA; [Link] Osmosis
(6) Esfahani, I. J.; Ataei, A.; Kim, M.; Kang, O.; Yoo, C. Parametric Plant [Link]
analysis and optimization of combined gas turbine and reverse osmosis (29) MacHarg, J. P.; McClellan, S. A. Pressure Exchanger Helps
system using refrigeration cycle. Desalin. Water Treat. 2012, 43, 149− Reduce Energy Costs in Brackish Water RO System. J. AWWA 2004,
158. 96, 44−47.
(7) Altaee, A. Computational model for estimating reverse osmosis (30) Redondo, J.; Busch, M.; Witte, J. D. Boron removal from
system design and performance: Part-one binary feed solution. seawater using FILMTEC high rejection SWRO membranes.
Desalination 2012, 291, 101−105. Desalination 2003, 156, 229−238.
(8) Yechiel, A.; Shevah, Y. Optimization of energy costs for SWRO (31) Hassan, A. M.; Al-Jarrah, S.; Al-Lohabi, T.; Al-Hamdan, A.;
desalination plants. Desalin. Water Treat. 2012, 46, 304−311. Bakheet, L. M.; Amri, M. M. I. Performance Evaluation of SWCC
(9) Feinberg, B. J.; Ramon, G. Z.; Hoek, E. M. V. Thermodynamic SWRO Plants. In Proceedings of IDA World Conference; Washington,
Analysis of Osmotic Energy Recovery at a Reverse Osmosis D.C. ; International Desalination Association: Topsfield, MA, 1991; pp
Desalination Plant. Environ. Sci. Technol. 2013, 47, 2982−2989. 208−224
(10) Bartman, A.; Christofides, P. D.; Cohen, Y. Nonlinear Model- (32) Hassan, A. M.; Abanmy, A. M.; Al-Thobiety, M.; Mani, T.; Al-
based control of an experimental reverse osmosis water desalination Luhibi, T.; Al-Masudi, I.; Al-Gherier, A. A.; Bakheet, L.; Amri, M. M. I.;
system. Ind. Eng. Chem. Res. 2009, 48, 6126−6136. Atiya, K. Performance Evaluation of SWCC SWRO Plants Part II. In
(11) Bartman, A.; Zhu, A.; Christofides, P. D.; Cohen, Y. Minimizing Proceedings of IDA World Conference; Washington, D.C. ; International
energy consumption in reverse osmosis membrane desalination using Desalination Association: Topsfield, MA, 1991; pp 412−439
optimization-based control. J. Process Control 2010, 20, 1261−1269. (33) Wilf, M.; Klinko, K. Effective new pretreatment for seawater
(12) Li, M. Minimization of energy in reverse osmosis water reverse osmosis systems. Desalination 1998, 117, 323−331.
desalination using constrained nonlinear optimization. Ind. Eng. Chem. (34) Tarquin, A. J. Desalination and Water Purification Research and
Res. 2010, 49, 1822−1831. Development Program Report No. 108: Volume Reduction of High-Silica
(13) Li, M. Reducing specific energy consumption in reverse osmosis RO Concentrate Using Membranes and and Lime Treatment; ,U.S.
(RO) water desalination: An analysis from first principles. Desalination Department of the Interior Bureau of Reclamation: Denver, CO, 2005.
2011, 276, 128−135. (35) Zhu, A. Energy and Cost Optimization of Reverse Osmosis
(14) Li, M. Optimal Plant Operation of Brackish Water Reverse Desalination. Ph.D. thesis, UCLA, Los Angeles, 2012.
Osmosis (BWRO) Desalination. Desalination 2012, 293, 61−68. (36) Kim, S.; Hoek, E. M. V. Modeling concentration polarization in
(15) Li, M. Optimization of multitrain brackish water reverse osmosis reverse osmosis processes. Desalination 2005, 186, 111−128.
(BWRO) desalination. Ind. Eng. Chem. Res. 2012, 51, 3732−3739. (37) Kurihara, M.; Yamamura, H.; Nakanishi, T. High recovery/high
(16) Li, M.; Noh, B. Validation of model-based optimization of pressure membranes for brine conversion SWRO process develop-
brackish water reverse osmosis (BWRO) plant operation. Desalination ment and its performance data. Desalination 1999, 125, 9−15.
2012, 304, 20−24. (38) Lemes, R.; Talavera, J. L. P.; Falcon, R.; Arocha, R.; Curbelo, J.;
(17) FILMTEC Membranes System Design: Multi-Stage System; The Platas, V.; Lorenzo, L. D.; Zarzo, D. Evolution of production and
Dow Chemical Company: Midland, MI; Form No. 609−02050−604. energy savings in SWRO plant of Las Palmas III. In Proceedings of IDA
(18) Franks, R.; Wilf, M.; Voutchkov, N.; Murkute, P.; Kizer, J. A World Conference; Taipei, Taiwan ; International Desalination
Pilot Study Using Seawater Reverse Osmosis Membranes in Combination Association: Topsfield, MA, 2011.
with Various Pretreatments to Meet the Challenges of Pacific Seawater (39) Penate, B.; de la Fuente, J. A.; Barreto, M. Operation of the RO
Desalination; Technical Report, Hydranautics: Oceanside, CA. kinetic energy recovery system: Description and real experiences.
(19) Schneider, B. Selection, operation and control of a work Desalination 2010, 252, 179−185.
exchanger energy recovery system based on the Singapore project. (40) MacHarg, J. P. Energy Optimization of Brackish Groundwater
Desalination 2005, 184, 197−210. Reverse Osmosis; Texas Water Development Board Contract Report
(20) Stover, R. L. Seawater reverse osmosis with isobaric energy Number 0804830845 ; Texas Water Development Board: Austin, TX,
recovery devices. Desalination 2007, 203, 168−175. 2011.
(21) Guirguis, M. J. Energy Recovery Devices in Seawater Reverse
Osmosis Desalination Plants with Emphasis on Efficiency and
Economical Analysis of Isobaric versus Centrifugal Devices. [Link].
thesis, University of South Florida, Tampa, FL, 2011.
(22) Ayyash, Y.; Imai, H.; Yamada, T.; Fukuda, T.; Yanaga, Y.;
Taniyama, T. Performance of reverse osmosis membrane in Jeddah
Phase I plant. Desalination 1994, 96, 215−224.
(23) Kammourie, N.; Dajanib, T. F. F.; Cioffib, S.; Rybarb, S.
SAWACO − North Obhor SWRO Plant Operational Experience.
Desalination 2008, 221, 101−106.
(24) Sanz, M. A.; Stover, R. L. Low Energy Consumption in the
Perth Seawater Desalination Plant. In Proceedings of IDA World
Congress; Maspalomas, Gran Canaria, Spain ; International Desalina-
tion Association: Topsfield, MA, 2007
(25) Dundorf, S.; MacHarg, J.; Sessions, B.; Seacord, T. F.
Optimizing Lower Energy Seawater Desalination. In Proceedings of
IDA World Conference; Dubai, United Arab Emirates ; International
Desalination Association: Topsfield, MA, 2009.
(26) Mohsen, M. S.; Gammoh, S. Performance evaluation of reverse
osmosis desalination plant: A case study of Wadi Ma’in, Zara and
Mujib Plant. Desalin. Water Treat. 2010, 14, 265−272.
(27) Yassin, A. Performance Evaluation of Five Years Operation
Experience of WMZM RO Desalination Plant; Taylor & Francis:
Abingdon, U.K., 2012.

17248 [Link]/10.1021/ie4031083 | Ind. Eng. Chem. Res. 2013, 52, 17241−17248