A closed-loop forward osmosis-nanofiltration hybrid system: Understanding process implications through full-scale simulation

A closed-loop forward osmosis-nanofiltration hybrid system: Understanding process implications through full-scale simulation

DES-13143; No of Pages 10 Desalination xxx (2016) xxx–xxx Contents lists available at ScienceDirect Desalination journal homepage: www.elsevier.com/...

2MB Sizes 0 Downloads 13 Views

DES-13143; No of Pages 10 Desalination xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

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

A closed-loop forward osmosis-nanofiltration hybrid system: Understanding process implications through full-scale simulation Sherub Phuntsho a,⁎, Jung Eun Kim a, Seungkwan Hong b, Noreddine Ghaffour c, TorOve Leiknes c, Joon Yong Choi d, Ho Kyong Shon a,⁎ a

Centre for Technology in Water and Wastewater, School of Civil and Environmental Engineering, University of Technology Sydney (UTS), 15 Broadway, Ultimo, NSW 2007, Australia School of Civil, Environmental & Architectural Engineering, Korea University, 1, 5-ka, Anam-Dong, Sungbuk-Gu, Seoul 136-713, Republic of Korea King Abdullah University of Science and Technology (KAUST), Water Desalination and Reuse Center (WDRC), Biological and Environmental Sciences & Engineering Division (BESE), Thuwal, 23955-6900, Saudi Arabia d Hyorim Industries Inc., Yatap-dong, Bundang-gu, Seongnam-City, 513-2, Gyeonggi-do, Republic of Korea b c

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Initial DS flow rate and concentration cannot be set at any arbitrary values. • Initial DS flow rate and concentration vary inversely for a fixed plant capacity. • Net DS mass flow rate mD is the most important parameter for a closed system. • mD is constant for a fixed plant capacity but increases with capacity and feed TDS. • FO and NF rejection rates influence feed solute accumulation in the closed system.

a r t i c l e

i n f o

Article history: Received 13 August 2016 Received in revised form 9 December 2016 Accepted 14 December 2016 Available online xxxx Keywords: Forward osmosis Nanofiltration Fertigation Simulation Desalination

a b s t r a c t This study presents simulation of a closed-loop forward osmosis (FO)-nanofiltration (NF) hybrid system using fertiliser draw solution (DS) based on thermodynamic mass balance in a full-scale system neglecting the non-idealities such as finite membrane area that may exist in a real process. The simulation shows that the DS input parameters such as initial concentrations and its flow rates cannot be arbitrarily selected for a plant with defined volume output. For a fixed FO-NF plant capacity and feed concentration, the required initial DS flow rate varies inversely with the initial DS concentration or vice-versa. The net DS mass flow rate, a parameter constant for a fixed plant capacity but that increases linearly with the plant capacity and feed concentration, is the most important operational parameter of a closed-loop system. Increasing either of them or both increases the mass flow rate to the system directly affecting the final concentration of the diluted DS with direct energy implications to the NF process. Besides, the initial DS concentration and flow rates are also limited by the optimum recovery rates at which NF process can be operated which otherwise also have direct implications to the NF energy. This simulation also presents quantitative analysis of the reverse diffusion of fertiliser nutrients towards feed brine and the gradual accumulation of feed solutes within the closed system. © 2016 Elsevier B.V. All rights reserved.

⁎ Corresponding authors. E-mail addresses: [email protected] (S. Phuntsho), [email protected] (H.K. Shon).

http://dx.doi.org/10.1016/j.desal.2016.12.010 0011-9164/© 2016 Elsevier B.V. All rights reserved.

Please cite this article as: S. Phuntsho, et al., A closed-loop forward osmosis-nanofiltration hybrid system: Understanding process implications through full-scale simulation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.12.010

2

S. Phuntsho et al. / Desalination xxx (2016) xxx–xxx

1. Introduction Desalination is one of the most reliable solutions to fresh water scarcity problems, however, the existing desalination technologies are still highly capital and energy intensive processes [1,2] making desalination not cost-effective in general for irrigation. Recent efforts have been made to develop less capital and energy intensive desalination technologies such as forward osmosis (FO), applied to a wide range of applications. Unlike conventional desalination based on the reverse osmosis (RO) process, the FO process uses a natural osmotic driving force created by the concentration difference between the highly concentrated draw solution (DS) and the feed solution (FS) when separated by a semipermeable membrane. The concentrated DS is diluted during operation, enabling it to be used either directly if suitable or processed further to separate pure water from the DS. However, producing pure water from the diluted DS requires a separation process such as a FOhybrid process [3]. One of the common approaches suggested is using established membrane based processes such as RO [4] and/or nanofiltration (NF) [5] as post-treatment processes to separate and regenerate draw solutes for potable water desalination. Based on recent studies, FO-hybrid systems could be more effective when the water treatment involves highly challenging waters such as highly impaired water sources from drilling flow back water [6,7], produced water from oil and gas extraction [8], brine treatment [9], raw sewage treatment [10], and aerobic osmotic membrane bioreactor [11, 12]. Fertiliser driven FO (FDFO) process is seen as a practical application of the FO process [13,14]. The diluted fertiliser DS from the FO process, which contains essential nutrients for plants, can be used for fertigation of crops if the nutrient concentration in the diluted DS meets required standards. The final diluted fertiliser concentrations that come out of the FO process, however, depend on the concentration level of the feed water at which they reach osmotic equilibrium [15]. This becomes a challenge when feed water sources with higher salinities are used, where the diluted fertiliser DS concentration does not meet the maximum nutrient concentration levels for direct fertigation. An easy solution is further dilution using available freshwater sources; however, this option is only practical when fresh water sources are available. The second option to dilute the fertiliser concentrations beyond osmotic equilibrium is a pressure assisted FDFO process to generate additional flux that could help further dilution of the fertiliser DS [16]. Another option studied by this group has been the FDFO-NF hybrid system where NF is used as a post-treatment to reduce the fertiliser concentration by removing the excess fertiliser from the diluted fertiliser DS and recovering it for further reuse [17]. The NF permeate, which is a more diluted fertiliser solution can then be applied for direct fertigation of crops. Recently, the FDFO-NF system has been studied at a pilot-scale level by our group [13,18]. The main objective of this study is therefore to conduct simulation of a full-scale FDFO-NF hybrid system to study how the various process parameters in the hybrid system affect each other when operated in a continuous closed-loop system based on a simple mass balance of the volumetric flows, draw solutes and the feed solutes. This study provides an enhanced understanding of some of the options and limitations of operating a full-scale FDFO-NF hybrid or in any FO-NF or FO-RO in a continuous closed-loop system. This study however does not include the process and energy efficiencies of the FO-NF hybrid system, as this particular area is being considered in a separate study. 2. Mass balance simulation of a continuous closed-loop FO-NF system The summary of the mass balance for the two types of flows (DS & FS), draw solutes and feed solutes for a full-scale FDFO-NF hybrid system is presented in Fig. 1. The simulation is entirely based on the ideal mass balance model in a closed system and which occurs independent

Fig. 1. Schematic layout of a continuous closed-loop FO-NF hybrid system considering the mass balance for the two flow rates, draw solutes and the feed solutes. For simulation mass concentrations (C) are measured in g L−1, flow rate or capacity Q in m3 day−1, rejection (R) and recovery rates (RR) in %.

of the process performances such as water flux (assuming infinite membrane area) including performance mechanisms. Therefore, the simulation models used in this study, especially related to the mass balance on the draw solutes and feed solutes, do not take into consideration the non-idealities that may exist in a real process and therefore assuming FO or NF modules as a black box. The only membrane properties accounted in this simulation models are the rejection rates and recovery rates of both the FO and NF membranes. Reasonable fixed rejection rates of the FO and NF membranes have been assumed for simulation although in reality, these rejection rates could slightly vary depending on its operating conditions under which the membrane is subjected. 2.1. Mass balance for the solution flows Considering a steady flow with no net build-up of volume in a continuous closed-loop FO-NF hybrid system (Fig. 1), the FO permeate and the NF permeate flow rates should be the same as the FO-NF plant capacity Qp which can be expressed as follows: Q p ¼ RRFO Q F;in

ð1Þ

where QF,in and RRFO are the inlet feed flow rates and feed recovery rate of the FO unit. If QD,in is the initial DS flow rate at the FO inlet, then the diluted DS flow rate at the FO outlet becomes (QD,in + Qp). The initial DS flow rate, however, is the same as the NF concentrate flow rate (i.e. no net accumulation in the system), referred to as the recycled DS in this study, and is given as a function of NF recovery (RRNF) rate as follows:  Q D;in ¼ Q p

 1 −1 RRNF

ð2Þ

2.2. Mass balance for the draw solutes Solute transfer in a FO process occurs in both the directions: forward diffusion of feed solutes and reverse diffusion of draw solutes. The reverse solute flux (RSF) measures the rate of reverse diffusion of draw solutes towards FS during the FO process. Since both RSF and water flux change with the DS concentration, the rate of reverse diffusion is

Please cite this article as: S. Phuntsho, et al., A closed-loop forward osmosis-nanofiltration hybrid system: Understanding process implications through full-scale simulation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.12.010

S. Phuntsho et al. / Desalination xxx (2016) xxx–xxx

3

also measured in terms of specific reverse solute flux (SRSF), defined as the ratio of the RSF to the water flux. SRSF actually measures the rate of mass of draw solute that diffuses reversely per unit volume of the water permeated through the FO membrane. For a particular FO membrane and draw solute, the SRSF has been observed to be generally constant irrespective of the DS concentration used [19]. The other term used is the inverse of the SRSF, known as the reverse flux selectivity (RFS) of the membrane which is defined as the rate of water extracted per unit mass of draw solute lost [19,20]. Higher RFS refers to better performing FO membrane with minimum loss of draw solutes during the FO operation. If CD,in is the initial DS concentration (mass per unit volume or in g L−1) at the FO inlet and SRSFD is the rate of draw solutes lost by reverse diffusion towards the FS, the following relationship can be obtained based on the mass balance of the draw solutes from Fig. 1:

processes have adequate or infinite membrane area for the process to occur until osmotic equilibrium or adequate feed NF recovery rates. Mass balance is basically a simple thermodynamic analysis which is in fact independent of the membrane flux. In a real world situation however, the membrane area will be finite and hence non-ideality of the process would exist which may affect the absolute values of the simulation outputs but not significantly on the output trend. If RNF,D is the draw solute rejection rate of the NF membrane, the concentration of the draw solute present on the NF permeate (CD,p) is given as follows:

 C D;in Q D;in ¼ C D;out Q D;in þ Q p þ SRSF D Q p

C D;out ¼ ð1−RRNF ÞC D;in −RRNF SRSF D

ð3Þ

where CD,out refers to the concentration of the diluted DS at the FO outlet, which then eventually becomes the feed to the NF process. Based on the principle of osmotic equilibrium, the DS cannot be diluted to a concentration below the osmotic pressure (π) of the FS [15, 21] without external intervention. This also means that the final diluted DS concentration (CD,out) therefore cannot go below the osmotic pressure of the incoming feed concentration (CF,in). Therefore, the following is true for any FO processes: πD;out ≥π F;in or C D;out ≥C F;in

ð4aÞ

Eq. (4a) is however true only for ideal condition (perfect rejection) without accounting the presence of feed solutes in the diluted DS accumulated as a result of the closed loop cycle and the fresh feed solute that came from the feed side each time. Eq. (4a) can be more accurately represented as follows that accounts the presence of the feed solutes:   πD;out þ π F;p2 ≥π F;in or C D;out þ C F;p2 ≥C F;in

ð4bÞ

where πF,p2 or CF,p2 refers to the feed solutes present in the diluted DS. More detail on this is covered under Section 2.3. The osmotic pressure is a function of concentration (C in mass per unit volume) and can be determined by van't Hoff equation π = nRTC /Mw for dilute or ideal solutions, where n, R, T and Mw are the van't Hoff factor, universal gas constant, absolute temperature and the molecular weight of the solute [15,22]. However, in most situations including in this study most electrolyte solutions are nonideal mixtures and hence corrections have to be made for calculating osmotic pressure by van't Hoff equation [23]. For the osmotically driven membrane processes, Wilson and Stewart recommend using molality and osmolality ideally derived from experimental measurements (vapour pressure, isopiestic, freezing point depression) when working with evaluating the FS and DS [22,24]. For higher concentrations, Pitzer correlations with Virial equations can be used to calculate the osmotic pressure [25]. However, these days, thermodynamic analysis software such as OLI Stream Analyser can be commonly used to estimate the osmotic pressure of the known solutes and hence it has been used in this study for convenience. If the DS and FS are made up of same solutes, osmotic equilibrium between the DS and FS would occur at the same solute concentrations, i.e. CF,in = CD,out. However, when the DS and FS are made up of different solutes say, the osmotic concentration does not occur at the solute concentrations and hence CF,in ≠ CD,out. For example, 5 g L− 1 NaCl has the same osmotic pressure (3.9 atm) as 9.5 g L− 1 (NH4)2SO4 as determined using thermodynamic analysis software OLI Stream Analyser (OLI Systems Inc., Morris Plains, NJ). Hence, CD,out in this study therefore must refer to the concentration of the diluted DS that exhibits equal osmotic pressure as the FS (CF,in). The mass balance is based on the assumptions of ideal models where the FO and NF

 C D;p ¼ 1−RNF;D C D;out

ð5Þ

Substituting Qp and QD,in in terms of RRNF, Eq. (3) can be further developed as follows: ð6Þ

The majority of the draw solutes are rejected by the NF membrane depending on the membrane rejection rate of the draw solute (RNF,D). Based on the mass balance at the NF module node, the following relationship is obtained:  C D;r Q D;in ¼ C D;out Q D;in þ Q p −C D;p Q p

ð7Þ

where CD,r refers to the draw solute concentration in the recycled DS (or NF concentrate) that goes back to the FO process. However, the value of CD,r is normally expected to be lower than the initial DS concentration (CD,in) at the DS inlet of the FO process due to loss by SRSFD and also permeation through the NF membrane. In order to maintain the same initial DS concentration (CD,in) or the driving force at the FO module inlet, the draw solute mass lost from the system must be externally replenished. The total draw solute replenishment rate mD,R (mass per unit time) can be represented as follows: mD;R ¼ Q p SRSF D þ C D;p



ð8Þ

As the lost draw solute by RSF reaches the FO feed brine, the concentration of the draw solutes (CD,brine) in the brine can be given as follows: C D;brine ¼

SRSF D Q p SRSF D ¼ 1 Q F;in −Q p −1 RRFO

ð9Þ

2.3. Mass balance for the feed solutes If RFO,F is the feed solute rejection rate of the FO membrane, the feed solute concentration in the FO permeate only (CF,p1) can be given by the following equation:  C F;p1 ¼ 1−RFO; F C F;in

ð10Þ

where CF,in is the concentration of the FS at the FO inlet. The actual feed solute concentration in the diluted DS (CF,p2a) at the FO module outlet is however slightly lower and is given as follows: C F;p2a ¼

Q p C F;p1 Q D;in þ Q p

ð11Þ

The total feed solutes present in the diluted DS should however be the sum of the feed solutes that constantly come together with the FO permeate (CF,p2a) and the feed solutes that already are present in the recycled DS (CF,p2b) except initially when a clean DS is used for the first time operation and hence CF,p2b(1) = 0. The number in the subscript within the bracket is used for identifying the number of cycles and the draw solute is recycled and reused in the continuous closed-loop FO-

Please cite this article as: S. Phuntsho, et al., A closed-loop forward osmosis-nanofiltration hybrid system: Understanding process implications through full-scale simulation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.12.010

4

S. Phuntsho et al. / Desalination xxx (2016) xxx–xxx

NF process. Hence, the total feed solute concentration in the diluted DS in the first cycle can be written as follows: C F;p2ð1Þ ¼ C F;p2a þ C F;p2bð1Þ ¼ C F;p2a ;

since C F;p2bð1Þ ¼ 0

ð12Þ

The feed solute concentration in the NF permeate (CF,p(1)) in the first cycle can be given as follows:  C F;pð1Þ ¼ 1−RNF; F C F;p2að1Þ

ð13Þ

where RNF,F is the NF feed solute rejection rate. The feed solute concentration in the recycled DS (or NF concentrate) returned to the FO process in the first cycle can be given as follows: C F;prð1Þ

Q p C F;p1 −C F;pð1Þ ¼ Q D;in

 ð14Þ

When this NF concentrate is recycled back to the FO process for the first time, some recycled feed solute is expected to return to the FO feed by reverse diffusion (SRSFF). This is because the feed solutes that are rejected by the NF membrane remain in solution as part of the recycled DS which ultimately renters the FO as part of the initial DS. It may be argued that the feed solute concentration of the FS is much higher than the feed solutes already present in the recycled DS and hence SRSFF may be less likely however this could be explained through the slight reduction in the feed solute rejection rate of the FO membrane. The feed solute already present in the recycled DS (CF,pr) slightly increases the feed solute of the FO permeate above as it adds to the CF,p1 that comes from the fresh FS and hence this slightly decreases the rejection rate of the FO membrane. However, for simplification, the SRSFF has been assumed to take place while the FO feed rejection rate (RFO,F) has been assumed to be constant. Since not all the feed solutes in the recycled DS would return to the FO feed, some of the feed solutes will re-enter the diluted DS in the second cycle. The accumulated feed solute concentration returned from the first cycle now present in the diluted DS as CF,p2b(2) can be given as follows: C F;p2bð2Þ ¼

C F;prð1Þ Q D;in −SRSF F Q p Q D;in þ Q p

ð15Þ

Hence, in the next or 2nd cycle, the total feed solute concentration in the diluted DS is given as follows: C F;p2ð2Þ ¼ C F;p2a þ C F;p2bð2Þ

ð16Þ

Since CF,p2(2) N CF,p2(1), there is a net accumulation of feed solute of CF,p2b(1) in the first full cycle of the DS in a closed FO-NF system. The value of CF,p2a is constant as long the FS and the FO membrane rejection rates remain constant. The increase in the total feed solute concentration in the diluted DS after reusing the first recycled DS will therefore subsequently affect the feed solute concentrations in the NF permeate, the recycled DS and the diluted DS in the next or second cycle. The feed solute concentrations in the NF permeate in the second cycle now becomes:  C F;pð2Þ ¼ 1−RNF; F C F;p2ð2Þ

ð17Þ

And the recycled DS or the NF concentrate becomes: C F;prð2Þ ¼

  Q D;in þ Q p C F;p2ð2Þ − 1−RNF; F Q p C F;p2ð2Þ Q D;in

ð18Þ

Here too, CF,p(2) N CF,p(1) and CF,pr(2) N CF,pr(1) due to accumulation of feed solutes within the closed system. From Eqs. (12) to (18), as the NF concentrate is continuously recycled and reused in the FO process as DS, the concentration of feed solutes will continue to rise as a result of continuous accumulation

within the closed-loop FO-NF system. If the number shown in the subscript bracket represents the number of times, the draw solute is recycled within the system, then the feed solute concentration in the diluted DS (CF,p2), NF permeate (CF,p) and the recycled DS (CF,pr) in the finite number of cycles (n) can be represented as follows: C F;p2ðnÞ ¼ C F;p2a þ C F;p2bðn−1Þ

ð19Þ

   C F;pðnÞ ¼ 1−RNF; F C F;p2ðnÞ ¼ 1−RNF; F C F;p2a þ C F;p2bðn−1Þ

ð20Þ



Q D;in þ RNF; F Q p C F;p2ðnÞ Q D;in   Q D;in þ RNF; F Q p C F;p2a þ C F;p2bðn−1Þ ¼ Q D;in

C F;prðnÞ ¼

ð21Þ

3. Results and discussion The simulation was carried out assuming (NH4)2SO4 (SOA) fertiliser as DS, and NaCl as FS. Although the simulation in this study is performed for FDFO-NF hybrid system using a specific draw and feed solutes, nevertheless, the simulation is applicable universally to any closed-loop FONF or FO-RO system using any type of DS and FS as long as the parameters are appropriately taken. The simulation was carried out using Microsoft Excel which can be accessible from the supplementary information. 3.1. Plant capacity, initial DS concentration and initial DS flow rate Unlike pressure based membrane processes where water flux can be increased by simply increasing the applied pressure, in the FO process however, the water flux can be increased mainly by increasing the DS concentration [26,27]. Fig. 2(a) shows how the required initial DS concentrations and the initial DS flow rates vary with the FDFO-NF plant capacity simulated using Eq. (3) for an ideal model assuming infinite membrane area. When a fixed initial DS flow rate (say 100 or 200 m3 d−1) is used, the required initial DS concentration increases linearly with the plant capacity. For the same plant capacity however this required initial DS concentration is lower when a higher initial DS flow rate of say 200 m3 d−1 compared to 100 m3 d−1. Likewise, a similar trend was also observed when a fixed initial SOA DS concentration is used. Fig. 2(b) further shows the variations between the initial DS concentrations and the required initial DS flow rates for a fixed plant capacity of 1000 m3 d−1. In general, the required initial DS concentrations and the initial DS flow rates vary inversely for a fixed plant capacity and FS. At the same initial DS concentration however, this required initial DS flow rate is higher for 20 g L−1 NaCl compared to 5 g L−1 NaCl FS further supporting the results in Fig. 2(a). A higher DS concentration is necessary for increasing the FO water flux however the results in the secondary axis of Fig. 2(b) show that, increasing the initial DS concentration also linearly increases the final concentration of the diluted DS at the FO outlet. The concentration of the diluted DS is a very important output parameter of the FO process because it directly affects the operating pressure and energy of the NF process. Achieving the lowest absolute concentration of the diluted DS (at osmotic equilibrium with FS) is not an option but essential which otherwise would impose significant energy penalty to the NF process. Based on the concept of osmotic equilibrium, a unit mass of DS has a fixed water extraction capacity or a draw solute carrying capacity which depends on the types of DS and the FS concentration used [15,28]. Hence, the higher values of initial DS concentrations (at fixed DS flow rates) or higher values of DS flow rates (at fixed DS concentrations) required for a larger plant capacity shown in Fig. 2(a) are because of the higher number of moles or the mass of the draw solutes required to extract increased volume of water. The water extraction capacity of the DS mass also decreases when a higher

Please cite this article as: S. Phuntsho, et al., A closed-loop forward osmosis-nanofiltration hybrid system: Understanding process implications through full-scale simulation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.12.010

S. Phuntsho et al. / Desalination xxx (2016) xxx–xxx

5

Fig. 2. Influence of FDFO-NF plant capacity and initial DS concentrations on various parameters. (a) Variations of the required initial DS concentration and initial DS flow rates with the plant capacity and (b) variations of the initial DS flow rates and the final diluted DS concentrations with the initial DS concentration. DS and FS concentrations (g/L) are obtained by multiplying mol/L concentration with the molecular mass of the solute.

FS concentration is used [15] and this explains why for a fixed plant capacity and DS concentration, a higher DS flow rate is required to increase the DS mass. For example, for a FDFO-NF plant capacity of 1000 m3 d− 1 and 5 g L−1 NaCl FS, the final diluted DS concentration expected is 13.2 g L−1 SOA at an initial DS flow rate of 100 m3 d−1 and initial DS concentration of 150 g L−1. The diluted DS concentration further increases to 24.6 g L−1 SOA if the initial DS flow rate is further increased to 200 m3 d−1 at the same initial DS concentration of 150 g L−1. These final diluted DS concentration values are much higher than the needed/target diluted DS concentration of 9.6 g L− 1 SOA that needs to be achieved at osmotic equilibrium with 5 g L−1 NaCl FS (osmotic pressure calculated using OLI Stream Analyser). The calculation shows that for a plant capacity of 1000 m3 d−1 and FS of 5 g L−1, the actual input parameters must be either 100 m3 d−1 initial DS flow rates with 111 g L−1 SOA initial DS concentration or 200 m3 d− 1 initial DS flow rates with 60 g L−1 SOA initial DS concentration in order to achieve the desired diluted DS concentration of 9.6 g L−1 SOA. Hence, these results show that, the absolute values of the initial DS concentration and its flow rate cannot be arbitrarily selected for a plant with a fixed volumetric capacity as they are influenced by the plant capacity and the FS concentrations which otherwise would result in a trade-off with the final concentration of the diluted DS affecting the NF energy consumption. 3.2. Draw solution mass flow rate in a continuous closed-loop FO-NF hybrid system The total volume of water a particular DS can extract depends on its total mass used, molecular weight, solubility and its equivalent concentration (at osmotic equilibrium) with the FS [15]. Hence, it may be more significant to relate the input parameters to the DS mass rather than the concentration and its flow rate. Since the initial DS concentration and initial DS flow rate are inversely related and their product is nothing but the mass flow rate, increasing the initial DS flow rate or DS concentration or both in fact increases the total DS mass flow rate (MD,in = QD,in × CD,in) supplied at the FO inlet. It was observed that for a fixed plant capacity, the total DS mass flow rate MD,in was not constant but increased linearly with the initial DS flow rate as presented in Fig. 3(a). However, by rearranging Eq. (3), the following relationship was observed constant for a fixed plant capacity and the FS:   Q D;in C D;in −C D;out ¼ Q p C D;out þ SRSF D ¼ mD;in ¼Constant

ð22Þ

For a fixed FO-NF plant capacity (Qp) and FS concentration (CF,in), the values of CD,out and SRSFD parameters are constant and hence the term

in the left hand side of Eq. (3), QD,in (CD,in − CD,out) should also be constant. The term QD,in (CD,in − CD,out) is defined in this study as the net DS mass flow rate (mD,in) of the FO process. It is observed that the required net DS mass flow rate (mD,in) is constant for a fixed plant capacity and FS concentration irrespective of the values of the initial DS flow rate or the initial DS concentration used for the system as shown in Fig. 3(a) secondary axis. This required net mass flow rate is higher for a FS at 20 g L−1 compared to 5 g L−1 FS. For example, for a plant capacity of 1000 m3 d−1 and 5 g L−1 NaCl FS, the net DS mass flow rate required is 10.1 ton d−1 SOA however this increases to 42.8 ton d−1 at a higher FS of 20 g L− 1 NaCl for the same plant capacity. If the net mass flow rate of 42.8 ton d−1 is used instead of 10.1 ton d−1 for a plant capacity of 1000 m3 d−1 and 5 g L−1 FS, the final diluted SOA DS concentration will be 42.3 g L−1 SOA instead of the desired 9.6 g−1 SOA because the permeate flow rate is not adequate to fully dilute the DS to the lowest required concentration significantly increasing the operating pressure and hence the energy consumption of the NF process. Fig. 3(b) further shows that, the required net DS mass flow rate increases with the FO-NF plant capacity and the FS concentrations. For example, for a plant capacity of 1000 m3 d−1, the net DS mass flow rates are 10.1, 20.6, 42.8 and 77.6 ton d−1 for FS of 5, 10, 20 and 35 g L−1, respectively. However, for a higher FO-NF plant capacity of 2000 m3 d−1, the required DS net mass flow rate doubles under the same FS conditions. The water extraction capacity per unit DS mass decreases at higher FS concentration and this is the reason why a higher DS mass is required for 20 g L−1 NaCl compared to 5 g L−1 NaCl. The higher DS net mass flow rate required at higher plant capacity is because, more water volume can be extracted only by increasing the mass supplied since the water extraction capacity of a unit mass of DS is fixed for a particular DS. The water extraction capacity of a unit mass of DS however decreases at higher feed concentration and hence more mass or net mass flow rate is required to extract the same volume of water from a feed with higher concentrations. If an excess DS mass is however loaded to the FO process for a given capacity and FS, the fixed FO permeate flow rate is not adequate to dilute the whole DS mass even though FO water flux would have increased. Hence, the concentrations of the diluted DS at the FO outlet increase due to the reduced DS dilution factor and which will have energy implications for the NF process. 3.3. NF feed recovery rates and the recycled DS concentrations Fig. 4 shows the variations of the recycled DS (NF concentrate) concentrations with the NF recovery rate simulated using Eq. (7). In a closed-loop FO-NF system, the initial DS concentration (CD,in) is equal to the recycled DS concentration (CD,R) with the addition of the total

Please cite this article as: S. Phuntsho, et al., A closed-loop forward osmosis-nanofiltration hybrid system: Understanding process implications through full-scale simulation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.12.010

6

S. Phuntsho et al. / Desalination xxx (2016) xxx–xxx

Fig. 3. Mass flow rates in a continuous and closed-loop FO-NF hybrid system. (a) Variations of the DS total mass flow rate and the net mass flow rate of the system with the initial DS flow rates and (b) variations of the DS net mass flow rates with the FO-NF plant capacity.

rate of DS mass lost from the system (mD,R) through reverse diffusion and NF permeation. The results in Fig. 4 indicate that, the NF process will have to operate at a much higher feed recovery rate to achieve a higher concentration of the recycled DS. For example, if the target recycled DS concentration is 100 g L−1, the NF has to operate at 91.3%, 81.6%, 60.3% and 25% for FS of 5, 10, 20 and 35 g L− 1, respectively. These recovery rates increase to 97.1%, 93.9%, 87.1% and 76.3% for FS of 5, 10, 20 and 35 g L−1 NaCl, respectively, if the target recycled DS concentration is increased to 300 g L−1. This suggests that, the initial DS concentration must be based on the maximum NF concentrate that can be achieved at optimum NF recovery rate which otherwise will significantly increase the NF operating pressure and hence the specific energy consumption. On the other hand, if a lower initial DS concentration is used, it reduces the recovery rate at which NF has to operate however this decreases the osmotic driving force which in turn lowers the water flux ultimately increasing the required membrane area and its cost. Therefore, in a continuous and closed-loop FDFO-NF system, the optimum initial DS flow rate and concentration must be determined based on the net DS mass flow rate for the system and the optimum operation of the NF process.

Fig. 4. Influence of the NF recovery rates on the concentration of the NF concentrate that is to be recycled as DS in the FO process.

3.4. Feed recovery rates and concentration of the lost draw solutes in the feed brine/concentrate Fig. 5(a) shows the rate at which the mass of the SOA draw solute could be lost due to reverse diffusion from the FDFO process of two different commercial 8040 FO membrane modules simulated using Eq. (8) but excluding DS lost from the system through NF permeation. The SRSFD data for CTA FO membrane is based on our earlier study [13]. Without the actual SRSFD data for the polyamide based thin film composite (TFC) 8040 FO membrane module, the SRSFD data obtained from labscale experimental study using TFC FO membrane (Toray Chemicals, Korea) is used for simulation. The CTA FO membranes with higher SRSF naturally have a higher rate of DS mass lost compared to the TFC FO membrane. The draw solute mass replacement rate due to reverse diffusion increases significantly with the increase in the plant capacity. For a plant capacity of 1000 m3 d−1 in Fig. 5(a), about 500 kg d−1 of SOA would be lost through reverse diffusion which needs to be replenished at an additional operating cost. These results therefore indicate that the cost of replenishing draw solute lost due to reverse diffusion for a large FDFO-NF plant using CTA FO membrane could be significant. This loss can be however significantly reduced to only 200 kg d−1 when a TFC FO membrane with higher selectivity is used. Besides economic costs, due to loss by reverse diffusion, discharging of FO brine containing NH4-N nutrient is also an environmental issue. Fig. 5(b) shows that the NH4-N concentration increases significantly with the feed recovery rate simulated using Eq. (9). Higher FO feed recovery rates translate to higher permeate flow rates and hence reduced FO brine flow rate, ultimately increasing the NH4-N concentration in the feed brine. In fact, Fig. 5(b) shows that the draw solute concentrations in the FO brine could increase exponentially with the increase in the FO feed recovery rate of above 90%. The concentration of fertiliser nutrients such as N and P in the FO feed brine is therefore likely to become an issue especially when the FO process is operated with high feed recovery rates. Brine management will not only require additional and complex treatment process but also incur additional costs. The environmental discharge of total nitrogen (TN) is 10 mg L−1 [29]. The concentrated brine from the FDFO plant could exceed the water quality standard for environmental discharge when the FDFO process is operated at feed recovery rates higher than 15% for CTA FO membranes and higher than 24% for TFC FO membranes. Besides, the reverse solute diffusion also has a significant impact on the membrane fouling [30,31] and scaling [32] due to complexation with the foulant or scaling species present in the feed and also resulting in cake enhanced osmotic pressure

Please cite this article as: S. Phuntsho, et al., A closed-loop forward osmosis-nanofiltration hybrid system: Understanding process implications through full-scale simulation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.12.010

S. Phuntsho et al. / Desalination xxx (2016) xxx–xxx

7

Fig. 5. Implication of the reverse diffusion of draw solutes towards the feed. (a) Variations of SOA DS mass replenishment rate lost due to reverse diffusion with the FDFO-NF plant capacity and (b) variations of the nitrogen nutrient (NH4) concentrations present in the FO feed brine for disposal for two types of FO membranes.

within the fouling layer [33]. The above results suggest the need for FO membranes with much higher reverse flux selectivity for commercial applications of the FDFO desalination plants.

3.5. Fixing the FO operational parameters in a continuous full-scale FO-NF hybrid system based on net mass flow rate The net DS mass flow rate in Fig. 3 and the recycled DS concentration in Fig. 4 are very useful for determining the operating parameters of a continuous full-scale FO-NF system. For example, for a plant capacity of 1000 m3 d−1, FS of 5 g L−1 NaCl and SRSFD of 0.5 g L−1 SOA, the required net DS mass flow rate is mD,in = Qp(CD,out + SRSFD) = 1000(9.6 + 0.5) = 10.1 ton d−1 SOA based on Eq. (22). The energy consumed by the NF process is critical to the FDFO-NF process and is influenced by its feed concentration and NF concentrate. Since the NF applied pressure and energy are directly related to NF concentrate and the feed recovery rates, a suitable value of the initial DS concentration (CD,in) must be first decided based on the optimum achievable recycled DS concentration CD,r values. Assuming an optimum CD,in of 100 g L−1 for the above example, the required initial DS flow rate is 101 m3 d−1 for the above calculated net DS mass flow rate. Hence, for a FDFO-NF plant capacity of 1000 m3 d−1 and FS of 5 g L−1 NaCl, using 100 g L−1 SOA as initial DS concentration will require an initial DS flow rate of 101 m3 d− 1. Any flow rate or DS concentration higher than this will only increase the concentration of the diluted SOA DS above 9.6 g L−1 increasing the NF energy consumption. Similar calculations can be performed for the same plant capacity but for a higher FS of say 20 g L−1 NaCl. In this case the mD,in increases to 42.8 ton d− 1 and with 100 g L− 1 SOA as initial DS will require an initial DS flow rate of 428 m3 d−1. From the above results and discussions, the net mass flow rate of the DS is one of the most important parameters for determining optimum operating parameters of a continuous closed-loop FDFO-NF hybrid system. The values of the input parameters cannot be arbitrarily selected when the plant capacity is fixed but determined based on the required DS net mass flow rate of the particular condition. Enhancing the driving force by only increasing the initial DS concentration for operating the FO process at higher water flux and recovery rate is not as straight-forward as increasing the driving force in the pressure based membrane process as it results in trade-off with the DS dilution factor. These results also clearly show that the DS flow rate cannot be simply increased to reduce the dilutive concentration polarisation effects or to increase mass transfer by increasing the crossflow velocity or Reynolds number as widely

suggested in the published literature through lab-scale studies [34– 36]. In a full-scale FO-NF system, the FO process therefore must rely on suitably designed spacers to enhance turbulence and mixing to lower the concentration polarisation effects. Process optimisation for a closed loop FO-NF system may therefore involve weighing the balance between the energy consumed by the NF process and the FO membrane area required which both depend on the initial DS concentration and its flow rate.

3.6. Accumulation of feed solutes in the draw solution of a closed FDFO-NF system Any differences in the rate of entry and exit of the feed solutes could result in unsteady state resulting in the accumulation of feed solutes within the closed FO-NF system. Fig. 6 shows the variations of the feed solute concentrations in the DS when the NF concentrate is recycled and reused a number of times in a continuous closed-loop FO-NF system simulated using Eqs. (11)–(21). As the DS is recycled several number of times, a build-up or net accumulation of feed solutes in the closed system is observed due to total feed solute mass entering the system being higher than exiting the system through the NF and FO membranes. The accumulated feed solute concentrations increase rapidly initially and then reaching a maximum steady concentration after

Fig. 6. Variations of feed solute concentrations at various points in the continuous closedloop FO-NF system. Simulation conditions are: CF,in = 20 g L−1 NaCl, RNF = 90%, RFO = 90%, Qp = 1000 m3 d−1, QD,in = 200 m3 d−1, SRSFF = 0.75 g L−1 NaCl.

Please cite this article as: S. Phuntsho, et al., A closed-loop forward osmosis-nanofiltration hybrid system: Understanding process implications through full-scale simulation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.12.010

8

S. Phuntsho et al. / Desalination xxx (2016) xxx–xxx

certain recycling times. For a FS of 20 g L−1 NaCl, 90% FO/NF rejection rates and SRSFF of 0.75 g L−1 NaCl, the accumulated feed solute concentrations in the recycled DS rapidly increase from zero (clean initial DS) to a steady maximum of 69 g L−1 NaCl after about 40 times recycling. The accumulated feed solute in the recycled DS is expected to positively contribute towards increasing the FO driving force and the water flux. However, this may likely affect the solubility of both the draw solutes and feed solutes in the NF concentrate and increase the potential for membrane scaling and flux decline in the NF process. Fig. 6 also shows (on secondary axis) a similar increase in the total feed solute concentrations in the diluted DS (CF,p2) as the DS is recycled several times reaching a steady concentration of 12.5 g L−1 NaCl. Unlike positive implication of the accumulated feed solutes in the recycled DS, in enhancing the FO driving force however, the accumulation of feed solutes in the diluted DS could be detrimental to the NF process. This is because the diluted DS after the FO process contains both draw solutes and the accumulated feed solutes and the total combined solutes (CD,out + CF,p2) will have osmotic pressure that exceeds the osmotic pressure of the FO FS. This means that, although the CD,out itself may have reached osmotic equilibrium with the FS (CF,in) but the combined solutes (CD,out + CF,p2) in the diluted DS will have osmotic pressure far greater than the initial FO FS and therefore result in energy penalty to the NF process which is not desirable. The feed solute concentration of the NF permeate from Fig. 6 is comparatively lower (1.5 g L−1 NaCl) however as this feed solute is present together with the draw solutes, the total solute concentration therefore should not exceed the salinity tolerance of the crops which otherwise further dilution becomes necessary. Besides, Na+ and Cl+ concentrations in the NF permeate should not exceed permissible limits for the crops. Fig. 7 shows the various factors responsible for influencing the accumulated feed solute concentrations. As shown in Fig. 7(a), when a FS of seawater quality is used (35 g L−1 NaCl), the accumulated feed solute concentrations in the recycled DS and diluted DS reach a maximum of 151 and 27.5 g L−1, respectively. Although such high concentrations of accumulated feed solutes in the recycled DS could definitely enhance driving force of the DS however, the increased osmotic pressure of the diluted DS could also significantly increase the required driving force and energy of the NF process. For example, for FO-NF plant with 35 g L−1 NaCl as FS, the diluted DS at osmotic equilibrium with the FS should be 77.1 g L− 1 SOA. Since this diluted DS contains both 77.1 g L−1 SOA and 27.5 g L−1 of accumulated feed solutes, the total solute concentration is 104.6 g L−1 with the accumulated feed solute making up almost 26.3% of the total solute mass. The osmotic pressure of the total solutes (104.6 g L− 1) in the diluted DS has been estimated at 53.6 atm (OLI Stream Analyser) which is almost twice as high as the

osmotic pressure of the original FS of 35 g L−1 NaCl (27.4 atm) which is a significant energy penalty to the NF process. If the feed solute accumulation is unavoidable, one of the options to avoid NF energy penalty could be to proportionately reduce the DS mass replenishment rate mD,R in Eq. (8) and let the accumulated feed solute compensate for the lost driving force or osmotic pressure from the initial DS as long as the NF permeate quality is maintained. The other option of course is to completely prevent the feed solute accumulation in the first place however this may not be possible with the polymeric membranes. The accumulation can be however significantly reduced by improving the selectivity of the FO and NF membranes. Fig. 7(b) presents the influence of feed solute rejection rates of FO and NF membranes on the variations of the maximum feed solute concentration in the NF permeate that is to be used for fertigation. At fixed NF rejection rates of 90% or 99%, the accumulated maximum feed solute concentration in the recycled DS decreases linearly with the increase in the FO membrane feed rejection rates (RFO,F) because of the decrease in the rate of entry of feed solutes in the closed system. On the other hand, at fixed feed solute rejection rates of the FO membrane (RFO,F of 90%, 95% and 97%), increase in the feed solute rejection rates of the NF membrane (RNF,F) does not show any change in the maximum accumulated feed solute concentrations in the NF permeate. The accumulated feed solute concentration in the NF permeate is however significantly lower at higher FO rejection rates as evident for 97% compared to 95 and 90%. From Fig. 7(b), it is therefore clear that, rejection rate of the FO membrane is the major factor responsible for the accumulated feed solute concentration in the NF permeate. The accumulated feed solute concentrations in the NF permeate can therefore be reduced by using high rejecting FO membranes and NF membranes. It is however well known that, improving the selectivity or rejection rate of the membrane also results in a trade-off with the water permeability which must also be taken into consideration for membrane development. Hence, this study underscores the importance of optimising the selectivity of the FO membrane for reducing the issues of feed solute accumulation and loss of draw solutes by reverse diffusion. It may be noted that the feed solute accumulation within the closed system was not affected by the plant parameters such as plant capacity, initial DS concentration and its flow rates. It is well known that, the feed solute rejection rates of the RO and NF membrane increase when operated at higher pressure or driving force. It is not sure whether a similar phenomenon can be applicable to the FO process when operated at different water fluxes at different DS concentrations. If this initial DS concentration affects the feed solute rejection rate of the FO membrane, it may also proportionately affect the feed accumulation rate as already evident from Fig. 7(b). This could be one interesting area of future experimental studies.

Fig. 7. (a) Variations of the accumulated feed solute concentrations in the recycled DS and NF permeate with the number of recycling times and (b) variations of maximum feed solute concentrations in the NF permeate at different FO/NF feed solute rejection.

Please cite this article as: S. Phuntsho, et al., A closed-loop forward osmosis-nanofiltration hybrid system: Understanding process implications through full-scale simulation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.12.010

S. Phuntsho et al. / Desalination xxx (2016) xxx–xxx

4. Conclusions The following conclusions are drawn from the mass balance simulation of a closed-loop full-scale FO-NF desalination system without accounting for the non-idealities of the system such as finite membrane area. • The initial DS flow rates and initial DS concentrations in a full-scale FO-NF hybrid system cannot be set at any arbitrary values for a FONF plant with a fixed volume output capacity. These parameters are directly influenced by the FO-NF plant capacity which otherwise increases the final concentrations of the diluted DS resulting in energy penalty to the NF process. For a fixed FO-NF plant capacity, the required initial DS concentrations and the initial DS flow rate are inversely related. • This study showed that the net DS mass flow rate is one of the most important parameters for the operation of a continuous closed-loop FO-NF system. The net DS mass flow rate is constant for a fixed plant capacity and feed condition however its value increases with increase in the plant capacity and the feed concentrations. • Reverse diffusion of fertiliser nutrients towards the feed brine could be one of the significant challenges of brine management in the FDFO process. The fertiliser nutrient concentration in the brine may easily exceed the local discharge standards especially when the FDFO is operated at higher feed recovery rates and this finding underscores the importance of having high reverse flux selectivity of the FO membranes. • The accumulation of feed solutes is likely to occur in a closed-loop FONF system, increasing rapidly with the number of DS recycling eventually reaching a steady state maximum concentration. Although the accumulated feed solutes in the recycled DS enhance the DS driving force and FO water flux however accumulation in the diluted DS poses a significant energy penalty to the NF process. The accumulation of feed solutes can be lowered by improving the FO membrane rejection rate although this could also reduce water permeability of the membrane. Hence, this study underscores the importance of having FO membrane with optimum selectivity. Symbols CF,in CD,in CD,out CD,p CD,SRF CD,r CF,SRF CF,p1 CF,p2 CF,p2a CF,p2b CF,p CF,pr MD,in mD,in mD,R (n) Qp QD,in QF,in RRNF RRFO RFO RNF,F

FO feed concentration initial DS concentration diluted DS concentration draw solute concentration in the NF permeate draw solute concentration in the FO feed brine recycled DS concentration feed solute concentration in the FO brine feed solute concentration in the FO permeate total feed solute concentration in the diluted DS feed solute concentration in the diluted DS from the FO permeate feed solute concentration in the diluted DS returned from the NF concentrate feed solute concentration in the NF permeate feed solute concentration in the recycled DS total DS mass flow rate net DS mass flow rate total DS mass replenishment rate the number of draw solute recycling times desalination plant capacity initial DS flow rate FS flow rate at the FO inlet NF feed recovery rate FO feed recovery rate feed salute rejection rate of the FO membrane feed solute rejection rate of the NF membrane

RNF,D SRSFD SRSFF

9

draw solute rejection rate of the NF membrane specific reverse solute flux of the draw solute specific reverse solute flux of the feed solute

Acknowledgements This research was supported under various funding: Industrial Facilities & Infrastructure Research Program (14IFIP-B087385-01) by the Ministry of Land, Infrastructure and Transport of the South Korean Government, King Abdullah University of Science and Technology (KAUST), Saudi Arabia, National Centre for Excellence in Desalination Australia (NCEDA), ARC Future Fellowship (FT140101208) and University of Technology Sydney chancellor's postdoctoral research fellowship. References [1] A. Subramani, M. Badruzzaman, J. Oppenheimer, J.G. Jacangelo, Energy minimization strategies and renewable energy utilization for desalination: a review, Water Res. 45 (2011) 1907–1920. [2] N. Ghaffour, T.M. Missimer, G.L. Amy, Technical review and evaluation of the economics of water desalination: current and future challenges for better water supply sustainability, Desalination 309 (2013) 197–207. [3] L. Chekli, S. Phuntsho, J.E. Kim, J. Kim, J.Y. Choi, J.-S. Choi, S. Kim, J.H. Kim, S. Hong, J. Sohn, H.K. Shon, A comprehensive review of hybrid forward osmosis systems: performance, applications and future prospects, J. Membr. Sci. 497 (2016) 430–449. [4] V. Yangali-Quintanilla, Z. Li, R. Valladares, Q. Li, G. Amy, Indirect desalination of Red Sea water with forward osmosis and low pressure reverse osmosis for water reuse, Desalination 280 (2011) 160–166. [5] C.H. Tan, H.Y. Ng, A novel hybrid forward osmosis – nanofiltration (FO-NF) process for seawater desalination: draw solution selection and system configuration, Desalin. Water Treat. 13 (2010) 356–361. [6] G. Chen, Z. Wang, L.D. Nghiem, X.-M. Li, M. Xie, B. Zhao, M. Zhang, J. Song, T. He, Treatment of shale gas drilling flowback fluids (SGDFs) by forward osmosis: membrane fouling and mitigation, Desalination 366 (2015) 113–120. [7] B.D. Coday, P. Xu, E.G. Beaudry, J. Herron, K. Lampi, N.T. Hancock, T.Y. Cath, The sweet spot of forward osmosis: treatment of produced water, drilling wastewater, and other complex and difficult liquid streams, Desalination 333 (2014) 23–35. [8] D.L. Shaffer, L.H. Arias Chavez, M. Ben-Sasson, S. Romero-Vargas Castrillón, N.Y. Yip, M. Elimelech, Desalination and reuse of high-salinity shale gas produced water: drivers, technologies, and future directions, Environ. Sci. Technol. 47 (2013) 9569–9583. [9] G. Chen, Z. Wang, X.-M. Li, J. Song, B. Zhao, S. Phuntsho, H.K. Shon, T. He, Concentrating underground brine by FO process: influence of membrane types and spacer on membrane scaling, Chem. Eng. J. 285 (2016) 92–100. [10] M. Xie, L.D. Nghiem, W.E. Price, M. Elimelech, A forward osmosis–membrane distillation hybrid process for direct sewer mining: system performance and limitations, Environ. Sci. Technol. 47 (2013) 13486–13493. [11] A. Achilli, T.Y. Cath, E.A. Marchand, A.E. Childress, The forward osmosis membrane bioreactor: a low fouling alternative to MBR processes, Desalination 239 (2009) 10–21. [12] R.W. Holloway, A.E. Childress, K.E. Dennett, T.Y. Cath, Forward osmosis for concentration of anaerobic digester centrate, Water Res. 41 (2007) 4005–4014. [13] S. Phuntsho, J.E. Kim, M.A.H. Johir, S. Hong, Z. Li, N. Ghaffour, T. Leiknes, H.K. Shon, Fertiliser drawn forward osmosis process: pilot-scale desalination of mine impaired water for fertigation, J. Membr. Sci. 508 (2016) 22–31. [14] L.A. Hoover, W.A. Phillip, A. Tiraferri, N.Y. Yip, M. Elimelech, Forward with osmosis: emerging applications for greater sustainability, Environ. Sci. Technol. 45 (2011) 9824–9830. [15] S. Phuntsho, S. Hong, M. Elimelech, H.K. Shon, Osmotic equilibrium in the forward osmosis process: modelling, experiments and implications for process performance, J. Membr. Sci. 453 (2014) 240–252. [16] S. Sahebi, S. Phuntsho, J.E. Kim, S.K. Hong, H.K. Shon, Pressure assisted fertiliser drawn osmosis process to enhance final dilution of the fertiliser draw solution beyond osmotic equilibrium, J. Membr. Sci. 481 (2015) 63–72. [17] S. Phuntsho, S. Hong, M. Elimelech, H.K. Shon, Forward osmosis desalination of brackish groundwater: meeting water quality requirements for fertigation by integrating nanofiltration, J. Membr. Sci. 436 (2013) 1–15. [18] J.E. Kim, S. Phuntsho, F. Lotfi, H.K. Shon, Investigation of pilot-scale 8040 FO membrane module under different operating conditions for brackish water desalination, Desalin. Water Treat. 53 (2015) 2782–2791. [19] W.A. Phillip, J.S. Yong, M. Elimelech, Reverse draw solute permeation in forward osmosis: modeling and experiments, Environ. Sci. Technol. 44 (2010) 5170–5176. [20] J.S. Yong, W.A. Phillip, M. Elimelech, Coupled reverse draw solute permeation and water flux in forward osmosis with neutral draw solutes, J. Membr. Sci. 392–393 (2012) 9–17. [21] S. Benavides, A.S. Oloriz, W.A. Phillip, Forward osmosis processes in the limit of osmotic equilibrium, Ind. Eng. Chem. Res. (2014). [22] A.D. Wilson, F.F. Stewart, Deriving osmotic pressures of draw solutes used in osmotically driven membrane processes, J. Membr. Sci. 431 (2013) 205–211.

Please cite this article as: S. Phuntsho, et al., A closed-loop forward osmosis-nanofiltration hybrid system: Understanding process implications through full-scale simulation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.12.010

10

S. Phuntsho et al. / Desalination xxx (2016) xxx–xxx

[23] L. Fitzsimons, B. Corcoran, P. Young, G. Foley, Exergy analysis of water purification and desalination: a study of exergy model approaches, Desalination 359 (2015) 212–224. [24] K.K. Reimund, J.R. McCutcheon, A.D. Wilson, Thermodynamic analysis of energy density in pressure retarded osmosis: the impact of solution volumes and costs, J. Membr. Sci. 487 (2015) 240–248. [25] K.H. Mistry, H.A. Hunter, J.H. Lienhard V, Effect of composition and nonideal solution behavior on desalination calculations for mixed electrolyte solutions with comparison to seawater, Desalination 318 (2013) 34–47. [26] 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 performance in a spiral wound module, J. Membr. Sci. 348 (2010) 298–309. [27] J.R. McCutcheon, R.L. McGinnis, M. Elimelech, Desalination by ammonia-carbon dioxide forward osmosis: influence of draw and feed solution concentrations on process performance, J. Membr. Sci. 278 (2006) 114–123. [28] A.D. Wilson, in: P. Wang (Ed.), Design of the Next-Generation FO Draw Solution, in Rational Design of Next-Generation Nanomaterials and Nanodevices for Water Applications, International Water Association (IWA) 2016, pp. 103–130.

[29] NSW-EPA, Protection of the Environment Operations (General) Regulation 2009. Version in 2009 No 211, 2009 (New South Wales State, Australia). [30] Q. She, R. Wang, A.G. Fane, C.Y. Tang, Membrane fouling in osmotically driven membrane processes: a review, J. Membr. Sci. 499 (2016) 201–233. [31] C.Y. Tang, Q. She, W.C.L. Lay, R. Wang, A.G. Fane, Coupled effects of internal concentration polarization and fouling on flux behavior of forward osmosis membranes during humic acid filtration, J. Membr. Sci. 354 (2010) 123–133. [32] S. Phuntsho, F. Lotfi, S. Hong, D.L. Shaffer, M. Elimelech, H.K. Shon, Membrane scaling and flux decline during fertiliser-drawn forward osmosis desalination of brackish groundwater, Water Res. 57 (2014) 172–182. [33] C. Boo, S. Lee, M. Elimelech, Z. Meng, S. Hong, Colloidal fouling in forward osmosis: role of reverse salt diffusion, J. Membr. Sci. 390–391 (2012) 277–284. [34] B.R. Babu, N.K. Rastogi, K.S.M.S. Raghavarao, Effect of process parameters on transmembrane flux during direct osmosis, J. Membr. Sci. 280 (2006) 185–194. [35] W.C.L. Lay, J. Zhang, C. Tang, R. Wang, Y. Liu, A.G. Fane, Factors affecting flux performance of forward osmosis systems, J. Membr. Sci. 394–395 (2012) 151–168. [36] S. Zhao, L. Zou, C.Y. Tang, D. Mulcahy, Recent developments in forward osmosis: opportunities and challenges, J. Membr. Sci. 396 (2012) 1–21.

Please cite this article as: S. Phuntsho, et al., A closed-loop forward osmosis-nanofiltration hybrid system: Understanding process implications through full-scale simulation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.12.010