Jiayou Xu ,Hongyu Wu ,Zhi Wang ,Zhihua Qiao ,Song Zhao ,Jixiao Wang

1 Chemical Engineering Research Center,School of Chemical Engineering and Technology,Tianjin University,Tianjin 300350,China

2 Tianjin Key Laboratory of Membrane Science and Desalination Technology,State Key Laboratory of Chemical Engineering,Collaborative Innovation Center of Chemical Science and Engineering,Tianjin University,Tianjin 300350,China

Keywords:Membrane-based process systems Separation Flue gas CO2 capture Biogas upgrading Natural gas

A B S T R A C T Membrane separation technology has popularized rapidly and attracts much interest in gas industry as a promising sort of newly chemical separation unit operation.In this paper,recent advances on membrane processes for CO2 separation are reviewed.The researches indicate that the optimization of operating process designs could improve the separation performance,reduce the energy consumption and decrease the cost of membrane separation systems.With the improvement of membrane materials recently,membrane processes are beginning to be competitive enough for CO2 separation,especially for postcombustion CO2 capture,biogas upgrading and natural gas carbon dioxide removal,compared with the traditional separation methods.We summarize the needs and most promising research directions for membrane processes for CO2 separation in current and future membrane applications.As the time goes by,novel membrane materials developed according to the requirement proposed by process optimization with increased selectivity and/or permeance will accelerate the industrialization of membrane process in the near future.Based on the data collected in a pilot scale test,more effort could be made on the optimization of membrane separation processes.This work would open up a new horizon for CO2 separation/Capture on Carbon Capture Utilization and Storage(CCUS).

1.Introduction

In recent years,membrane separation technology has received more and more attentions,since membrane process is energy efficient,environmentally friendly and easy to scale up.With the improvement of membrane selectivity and permeance[1,2],the feasibility of membrane process development for CO2separation is increasing.The advantages and disadvantages of different CO2-separation processes were summarized in Table 1.This work mainly reviews the membrane processes for CO2separation,including post-combustion CO2capture,biogas upgrading and natural gas carbon dioxide removal.

As known,membrane performance is mainly characterized by two parameters[1]permeance and selectivity.Although membranes with extremely high permselectivity are available for different processes,the feasibility of membrane process depends on not only the membrane selectivity and permeance,but also the operating conditions(including operating pressure and number of stages).Therefore,membrane-based CO2separation processes discussed in this work are intending to illustrate the membrane process for specific application based on process simulation and economic cost estimation.The trade-off between membrane permeance and selectivity always exists,and it could be described by the empirical Robeson upper bound[5,6]for CO2versus different gases.Therefore,a good design of membrane separation processes has to take the trade-off between permeance and selectivity into consideration.

Mathematical modeling and experiment testing are two main aspects for the investigation of membrane-based gas separation processes.There have been many researchers[7–12]developing and improving the mathematical model,since Weller and Steiner[7]developed the mathematical model for binary component gas separation membrane process.Solutions for multicomponent membrane process were presented by Shindo et al.[10]who investigated a membrane process operated with three commonly reported flow patterns(including cross flow,counter-current flow and co-current flow).The differences among these three flow patterns can be negligible when the stage cut(the ratio of the permeate flow rate to the feed flow rate)is relatively low(around 0.2).The product purity calculated by cross flow is between counter-current flow and co-current flow[13].Therefore,the cross flow pattern is recommended for the calculation of membraneprocesses,especially for the simulation of spiral wound membrane.When simulating the separation performance of hollow fiber membrane,the counter-current flow pattern is often applied due to the high driving force along the membrane module[11].Recently,further improvement on membrane process simulation has been proposed.For instance,Alshehri et al.[12]have improved Shindo's model by adding minor revisions to take pressure drop,gas nonideal behavior,pressure and temperature dependence of membrane permeance and concentration polarization into consideration.Coker et al.[11]introduced a stage-wise mathematical model by dividing the length of the hollow fiber into many sections and solved the mass balances in each section.The developed model could be applied for co-current flow,counter-current flow,and cross flow.The effect of permeate sweep,pressure-dependent permeance and pressure drop also could be taken into consideration.The boundary layer on either side of the membrane was found to result in the decrease of concentration gradients across the membrane(i.e.concentration polarization)and thus the reduction of driving force.Scholz et al.[14]highlighted that when operated with high pressure,the Joule–Thomson effect,pressure drop and real gas behavior should be considered in the membrane process.Since the mathematical modeling validated experimentally is accurate enough,it offers the advantage of cost effectiveness,safety and flexibility to extensive parametric studies compared to experiment testing,and most of the researches with regard to membrane separation processes could be conducted by the modeling.

Table 1 The advantages and disadvantages of different CO2-separation processes[3,4]

In this paper,according to the significant differences of the specific separation system on feed pressure,feed gas CO2concentration and membrane permselectivity,we review three frequently reported and promising membrane processes for CO2separation,including the post-combustion CO2capture,biogas upgrading and natural gas carbon dioxide removal,and discuss some complementary developments of membrane process design.As for the other membrane processes for CO2separation,which behave relatively less reported and not systematic enough,they are not within the scope of this review.We guess that there might be a great bottleneck to be tackled before the prevailing studies of these specific membrane processes,that is,the membrane permselectivity of their materials waits for further enhancing and immune tolerance of the membrane should be investigated systematically for the complicated separating conditions.Besides,the membrane technology for CO2separation has not shown its enough maturity to extend to various industrial processes.Therefore,the reported researches mainly focus on several typical membrane processes,i.e.postcombustion CO2capture,biogas upgrading and natural gas carbon dioxide removal,which would be discussed in detail below.

2.Membrane-based Gas Separation Processes

2.1.Post-combustion CO2 capture from flue gas

The increased CO2concentration in environment leads to global warming.Excessive increase of atmosphere temperature is responsible for severe environmental problems including the rise of water-level in sea and the higher number of storms and floods[15].Coal-fired power plant is one of the main sources of CO2emission from fossil fuel.In order to mitigate the climate change due to CO2emission,CO2capture from flue gas has received considerable attentions.As a result,a great number of separation processes have been investigated intensively through laboratory test,simulation studies and even pilot scale test.

Membrane separation technology is one of the most promising technologies for CO2capture from flue gas,when taking into account the benefits of membrane processes,including simplicity of operation,modular construction,small footprint,and no hazardous by-product emissions.However,CO2capture from flue gas based on the membrane process is still not fully explored,since the separation targets proposed by the U.S.Department of Energy(DOE)with product purity higher than 95%and recovery higher than 90%are too difficult to achieve with low energy consumption and capture cost.The low CO2concentration and low pressure in flue gas are responsible for the difficulty of separation process,high energy consumption and capture cost.Many researchers[3,16–31]are optimizing membrane-based CO2capture processes from flue gas and intend to find the optimal materials and operating conditions with the lowest energy consumption and capture cost.

The low CO2concentration[28]in power plant flue gas(13%–15%for typical coal fired power plant)is one of the most important obstacles for post-combustion CO2capture.Favre[3]mentioned that with a feed gas CO2concentration of 10%,the energy consumption required for a membrane system was too high to be accepted,even with selectivity above 120.When the CO2concentration in feed gas exceeded 20%,the energy consumption for the membrane system decreased,which would make the membrane process more competitive than the traditional absorption process.For a typical case,with a selectivity of 60 and a downstream pressure of 3000 Pa,a single stage membrane system would generate an energy requirement of 0.7 MJ·kg-1,for a feed gas with 20%CO2.Therefore,the low CO2concentration in flue gas is the main bottleneck to realize a capture cost as low as 20 USD·t-1CO2.

For the single-stage membrane process shown in Fig.1,in order to realize the separation target of 95%purity and 90%recovery of CO2,high selectivity is needed.As reported by Favre[3],selectivity above 200 is required to realize the targets stated above.Further studies conducted by Yang et al.[25]point out that the minimal selectivity is a function of permeate/feed pressure ratio.The decrease of permeate/feed pressure ratio will lead to a decrease in selectivity required to realize the targets proposed by DOE.For instance,when the pressure ratio approaches 0,a selectivity of 300 is needed for the fulfillment of 95%CO2purity and 90%CO2recovery.While with a permeate/feed pressure ratio of 0.02,selectivity above 600 is required to realize the same product purity and recovery.Therefore,it is not practical for the single-stage membrane process to realize the targets proposed by DOE,mainly because the product purity is restricted by the low feed gas CO2concentration and the trade-off between product purity and recovery.

Relatively speaking,it is easier for the two-stage membrane process to realize the targets proposed by DOE,since the cycle gas significantly improves the CO2recovery.Fig.2 illustrates a typical two-stage membrane system with cycle gas.The first stage provides concentrated gas which is directed to the second stage to further improve its purity.In this case,the second stage residue stream is recirculated to the first stage to ensure the high CO2recovery.As reported by Yang et al.[25],with a membrane selectivity of 52,CO2permeance of 3.12×10-3m3(STP)·m-2·s-1·MPa-1and pressure ratio of 0.081,the separation target proposed by DOE could be satisfied by a two-stage membrane system with a capture cost of 48.7 USD·t-1CO2.If the permeate/feed pressure ratio decreased to 0.05,the two-stage membrane system could realize the separation target with membrane selectivity as low as 40.The membrane is operated at the temperature of 313 K.

Fig.1.Single-stage membrane processes for flue gas CO2 capture.

As shown in Fig.2,compared with the single-stage membrane process[29],the two-stage membrane process requires a larger number of compressors or vacuum pumps which consume more power,and the cycle gas brings about higher membrane area requirement.Therefore,the for two-stage membrane process,the reduction of energy consumption and membrane area is the main target of flue gas CO2capture process optimization.

Zhao et al.[28]found that the energy consumption effectively decreased with the improvement of membrane CO2/N2selectivity.With the pressure ratio of 10 in the first stage and pressure ratio of 4 in the second stage,by increasing membrane CO2/N2selectivity from 20 to 40,the energy consumption decreased more than halved.However,when the selectivity increased to higher value,the decreasing tendency became slower.Therefore,for the two-stage membrane process,a CO2/N2selectivity of 40 might bring about enough energetic advantage.The required membrane area decreased with the increase of CO2permeance,thus the capture cost also decreased accordingly.In sum,the improvement of membrane selectivity and permeance is contributive to the reduction of energy consumption and membrane area,respectively.When the Robeson upper bound[5,6]was imposed on membrane process design,a medium selectivity combined with relative high permeance was found to be more advantageous in improving postcombustion carbon capture membrane performance,than simply using a membrane with high selectivity.Similar conclusion was also summarized by Zhang and coworkers[27].They found that with the increase of CO2/N2selectivity,energy consumption decreased but a larger membrane area was required for the two-stage membrane process.Thus,a trade-off between energy consumption and membrane area was encountered.The optimal selectivity was found to range from 70 to 90 for the two-stage membrane process.

Fig.2.Two-stage membrane processes for flue gas CO2 capture.

In order to obtain the driving force for the membrane process,a vacuum pump in the permeate side or compressor in the feed side should be installed.Through the analysis conducted by Yang and coworkers[25],it was found that the vacuum process consumed less energy than pressurized process,but required higher membrane area.A higher pressure difference greatly would reduce the membrane area of the process,however,with a higher energy consumption.Actually,the success of membrane processes for CO2capture from flue gas depends not only on excellent membrane permselectivity but also on process operating conditions.The appropriate pressure on either the permeate side or feed side is crucial for the optimization of the membrane process.

Huang et al.[19]emphasized the importance of the pressure ratio on membrane separation performance.Pressure ratio not only limited the product purity but also led to the change of compressor capital cost and energy consumption.The balance between pressure ratio and selectivity should be highlighted by researchers afterwards.For a two-stage membrane process shown in Fig.2(b),by pulling a vacuum on the permeate side,a pressure ratio of 10 provided the minimal energy consumption.With this pressure ratio,membranes with high permeance are always preferred,but the optimal membrane process configuration might not need a membrane of the highest selectivity.The optimal selectivity of the process is decided by both the pressure ratio and the other operating conditions.

Recently,more concerns have been focused on multi-stage processes,with respect to energy consumption and CO2capture cost.For instance,Shao et al.[23]investigated a two-stage membrane system for CO2capture from flue gas with 13%CO2.As shown in Fig.3,the first stage(A1)was responsible for improving CO2concentration to 85%.The second stage(A3)was able to provide a permeate gas with CO2concentration above 99%.While A2 was meant to improve CO2recovery by capturing the CO2in residue gas and recycling it back to the first stage.The capture cost could be reduced with further improvement of membrane selectivity.With a CO2/N2selectivity of 200,membrane permeance of 100 GPU(1 GPU=10-6cm3·cm-2·s-1·cm Hg-1=3.346×10-10mol·m-2·s-1·Pa-1),feed pressure of 0.24 MPa and permeate side pressure of 14 kPa, capture cost as low as 17.9 USD·t-1CO2could be achieved with 24.4%output power consumption,which is more competitive than the traditional amine absorption process for CO2capture from coal fired flue gas.

Fig.3.The two-stage membrane separation for CO2 capture from the coal-fired flue gas designed by Shao and coworkers[23].

Different from the traditional studies on single-stage and two-stage membrane process performance,advanced algorithms take its place among the research methods,which could take more factors into consideration during the optimization of the membrane separation process.For instance,Yuan et al.[26]used a genetic algorithm to simultaneously optimize energy consumption and membrane area requirement for single-stage and two-stage membrane processes.The decision variables included the operating pressures,temperature and intermediate composition.The lowest energy consumption was found to be 1.1 GJ·t-1CO2for a hybrid two-stage process with a N2-selective membrane in the first stage and a CO2-selective membrane in the second stage.Arias et al.[31]constructed a superstructure which included several candidate configurations as a nonlinear programming(NLP)model to figure out the optimal number of stages,with lower membrane area and total cost,and their results showed that the objective product CO2concentration determined the optimal number of stages.With product CO2concentration ranging from 90%to 93%,the two-stage membrane system with a recycle stream was found to be the optimal configuration.With CO2concentration ranging from 94%to 96%,the two-stage membrane system and three-stage membrane system were found to remain a minimum of the total cost.More recently,this conclusion also has been confirmed by researchers afterwards.Gabrielli et al.[17]utilized a genetic algorithm to find the appropriate process configuration with minimum energy consumption and membrane area.With product purity fixed as 95%,the differences between two-stage and three-stage membrane processes of membrane area and energy consumption were found to be negligible.As such,the two-stage membrane system was proven to be the most practical design for the realization of separation targets proposed by DOE while the three-stage membrane system did not notably reduce the capture cost but remarkably increased the process complexity.For the CO2capture from flue gas,more focus should be put on the optimization of the two-stage membrane system in the future.

Fig.4.Simplified flow diagram of a two-step vacuum membrane process to capture CO2 in flue gas from a coal-fired power plant designed by Merkel and coworkers[21].The membrane with a CO2 permeance of 1000 GPU and a CO2/N2 selectivity of 50 was used in the calculations.

Another way to reduce the energy consumption of CO2capture from flue gas is the utilization of air as sweep gas.Merkel et al.[21]proposed a two-step membrane process by employing air as sweep gas to capture CO2from flue gas and recycling the sweep air back to the boiler.In the two-step membrane process,the residue gas leaving the first step is sent to the second step to improve the CO2recovery.While for the two-stage membrane process,the permeate gas of the first stage is sent to the second stage to improve the CO2concentration.Sweep operation was found to consume less energy than that with a compressor or vacuum pump,since the sweep gas could increase the driving force for CO2permeation without improving pressure.As shown in Fig.4,in the first step,a vacuum pump was used to provide the driving force for CO2separation.Since the CO2enriched permeate gas passing through the vacuum pump was only a fraction of the volume of the flue gas,the energy consumption caused by the vacuum pump was much lower than that consumed by compressing the flue gas.The CO2concentration in the first step residue gas was around 7%.In the second step,a counterflow/sweep configuration was applied to further reduce the residue gas CO2concentration to 1.8%by utilizing air as sweep gas on the permeate side.The CO2in the sweep air was returned back to the boiler to improve the flue gas CO2concentration to 20%.As a result of high feed gas CO2concentration and high residue gas CO2concentration,the first step permeate gas CO2concentration could be as high as 83%,which still required further treatment to improve the CO2concentration to 95%.Overall,with a product purity of 95%and CO2recovery of 90%,a capture cost of about 23 USD·t-1CO2could be realized.

The cost-sensitivity analysis conducted by Ramasubramanian et al.[21]also proved that the membrane-based air sweep process could realize the DOE capture targets.To further reduce the cost of the membrane process,the membrane with CO2/N2selectivity above 140 and a CO2permeance of 3000 GPU was recommended to realize a capture cost of 24 USD·t-1CO2.Higher CO2/N2selectivity significantly reduced energy requirement due to the higher CO2concentration in the first step.Increasing CO2permeance was definitely beneficial due to ensuing decrease in membrane area.Therefore,further improvement in membrane permselectivity was crucial to reduce the capture cost of the membrane-based air sweep process.

Franz et al.[16]carried out an energetic and economic analysis to compare the process with sweep gas and the process merely with a compressor.The result also showed that using sweep gas in a twostage membrane system led to lower energy consumption than using a compressor.However,due to the large membrane area accompanied,the cost of the membrane process using air sweep was nearly the same as that without sweep gas.Using sweep gas also resulted in the variation of feed gas O2concentration and the change of combustion temperature in the boiler.Therefore,further investigation was needed to prove the feasibility of this method.

Although the membrane-based separation processes have been proved to be a competitive solution to separate CO2from the mathematic point of view,more effort still should be put on the realization of the membrane process in a pilot scale or even in industrial scale.

During 2012–2014,a membrane-based CO2capture process from coal-fired power plant flue gas using the full-scale Polaris?module had been installed and investigated at the National Carbon Capture Center in Wilsonville,Alabama,USA[24].Both cross-flow and countercurrent sweep spiral-wound modules were employed to construct a two-step membrane process,which ran up to 1800 h and effectively provided an enriched permeate gas with a CO2concentration of 62.6%and CO2recovery around 90%.As shown in Fig.5,the sweep gas could reduce the first step residue gas CO2concentration to 2.78%,which effectively improved the CO2recovery.These results were in good agreement with the process design proposed by Merkel and coworkers[21].Hence the experiment gathered from this pilot plant could be potentially applied to the design and construction of a larger demonstration plant.

Recently, He et al. [18] reported a pilot scale test using polyvinylamine(PVAm)based fixed-site-carrier(FSC)hollow fiber membranes at the Tiller plant(Trondheim,Norway).As shown in Fig.6,a membrane system with two modules in parallel was constructed for this pilot scale test.With a feed flue gas CO2concentration of 9.5%,a product CO2purity of 55%could be obtained by a singlestage process.The water permeation through the membrane was tested by the condenser in Fig.6.The water permeance was found to be much higher than the CO2permeance,so it was necessary to maintain high water concentration in the feed gas in the industrial application.The results in this single-stage process could provide the necessary information for the design and optimization of a two stage process for CO2capture from flue gas.The authors indicated that pressure should be optimized,for the pressure significantly could affect the membrane permeance of facilitated transport membrane.As a result,technoeconomic analysis should be conducted to balance the required membrane area and the energy consumption.

Thus far,membrane separation process is becoming an important option for CO2capture from coal-fired power plant flue gas according to the recent researches.The choice of membrane materials and process configuration have been explored by many researchers.However,the methods reported are not systematic enough to provide the optimal solution to the process designer.Several important aspects still should be stressed deeply in the development of the membrane process listed below.

1.The separation targets proposed by DOE could be realized by the two-stage membrane process.The two-stage membrane process should be investigated systematically enough to explore its potential in reducing energy consumption and membrane area,before a new process with much lower energy consumption or capture cost was developed.

2.Membrane material development of lab scale should be tailored according to the requirement of optimal conditions in process design,considering the pressure ratio,membrane selectivity,membrane permeance,feed gas temperature and feed gas CO2concentration.

3.Compared to traditional chemical absorption technology, the membrane-based separation process is still underdeveloped.Therefore,the process design is important for the capture system optimization,and pilot scale and full scale tests are the ultimate measurement for the success of membrane-based CO2capture technology from flue gas.

4.The improvement of membrane selectivity and permeance contributes to a lower energy consumption and lower capture cost.For the improvement of membrane selectivity,for example,the polyethersulfone(PESU)membrane[32]with the CO2/N2selectivity of 34 would be potentially suitable,and the PEOT/PBT/ZIF-71 membrane[33]with the CO2/N2selectivity of 52.6 is also recommended for the post-combustion CO2capture.For the improvement of membrane permeance,the optimal PIM-CD/PDMS/PAN composite membrane[34]enjoys a CO2permeance of 483 GPU,and the PDMS/PAN composite hollow fiber membrane[35]with a CO2permeance of＞5000 GPU is reported.

Fig.5.Performance data for 1 TPD carbon capture membrane system reported by White and coworkers[24].1 psi=6894.76 Pa.

Fig.6.Membrane system tested at 23°C and a feed pressure of 2 bar with a feed flow controller set to 40 m3·h-1 flue gas,reported by He and coworkers[18].1 bar=0.1 MPa.

2.2.Biogas upgrading

Biogas is playing a more and more important role in the renewable energy market in recent decades.Produced from a variety of aspects,such as municipal sewage treatment waste,human or animal waste and farm agricultural waste,typical biogas mainly contains 55%–65%CH4,35%–45%CO2and other minor components(such as water vapor,hydrogen sulfide,volatile organic compounds and siloxanes)[36].Biogas is a common fuel for stoves,internal engines,gas turbines and fuel cells,and also could be injected into natural gas grids after purification[37].For the sake of increasing the calorific value of biogas,the unwanted components,such as CO2and H2S,should be removed from raw biogas through the biogas upgrading process.Besides,CO2capture from biogas can be applied for other applications,such as enhanced oil recovery(EOR)and algae production,which would further reduce the cost of biogas upgrading.Therefore,the biogas upgrading technology is attracting more and more interest in the industry processes.Depending on competitive energy consumption and operating costs,membranebased technology has been more and more extensively used in the biogas upgrading field.

Recently,the developments of membrane materials with high CO2/CH4selectivity and permeance have boosted the development of the membrane based biogas upgrading process[38].This section reviews the reported biogas upgrading process using a CO2separation membrane.The single-stage membrane process for biogas upgrading has been detailedly discussed using typical polymeric membranes.The trade-off between CH4recovery and CH4purity was observed,which is equivalent to a negative correlation between CH4product purity and CH4recovery[39].Even though the improvement of CO2/CH4selectivity contributes to higher product purity for a specific CH4recovery,the single-stage gas membrane system is still unable to provide a product with sufficiently high CH4purity and high CH4recovery simultaneously[40].Hence,multi-stage membrane systems are developed for biogas upgrading to tackle this drawback.

Fig.7.Schematic of(a)single-step membrane design and(b)two-step membrane design for biogas upgrading.

Havas and Lin[41]evaluated the potential of single-step and twostep membrane processes for biogas upgrading,as depicted in Fig.7.Systematic techno-economic analysis was performed to find out the optimal membrane material that could provide the optimal capital cost and operating cost of the membrane system.The membrane selectivity and permeance were determined by Robeson's upper bound[6]for a CO2/CH4separation membrane.For the two-step membrane process,a minimum operating expense was detected with CH4purity fixed as 96%and CH4recovery fixed as 85%.With the increase of CO2/CH4selectivity,the operating cost firstly decreased and then increased.Under the premise of relatively high CO2permeance above 2000 GPU,an optimal membrane with selectivity ranging from 10 to 25 could provide the lowest operating cost of 0.037 USD·m-3,when the membrane was operated with a pressure ratio of 10.When the optimal membrane confirmed,the cost of biogas upgrading is dominated by the compressor including the capital depreciation and utility expenses,which amounted to approximately 85%of the total cost.In addition,further decrease of membrane skid cost would lead to the decrease of operating cost and the increase of CO2/CH4selectivity with the minimum operating cost.In conclusion,the two-step membrane process with optimal selectivity provides lower operating cost than single-step membrane process.Membranes should have high CO2permeance and sufficient CO2/CH4selectivity(10–25).As such,the polyethersulfone(PESU)membrane[32]with a CO2/CH4selectivity of 35.5 is high enough for the two-step membrane processes for biogas upgrading.

Shao et al.[42]designed a two-stage membrane process to generate pipeline quality methane(97%purity)with 99%recovery.The advantages of membrane technology over pressure swing adsorption(PSA)in upgrading bio-methane from the biogas were demonstrated.The membrane selectivity was set as 55 and the CO2permeance was set as 25 GPU,with the operating temperature of 303.2 K.With the aim of reducing the overall process cost,membrane process design was optimized by the moderation of pressure at each side of the membrane.The first stage upstream/downstream pressure ratio was found to range from 8.5 to 8.7,and the second stage feed pressure was set to be 0.5 MPa. The separation cost of membrane process was 0.89 USD·MMBtu-1while the typical cost for PSA ranged from 5 USD·MMBtu-1to 7 USD·MMBtu-1.Moreover,when processing a 200 m3·h-1biogas stream,the energy consumption for the two-stage membrane process was 1.69 MJ·kg-1(equal to 1.21 MJ·m-3),which was lower than the energy consumed in the PSA process.The result proved that the membrane process was better than the traditional PSA adsorption process for producing pipeline quality methane considering methane recovery and processing cost.

In order to further reduce the energy consumption and operating cost,Valenti et al.[43]elaborately investigated single-stage and twostage membrane processes by utilizing cellulose acetate spiral wound membrane.The results showed that for each design,specific membrane area ranged from 1.1 to 2.4 m2·h·m-3,while specific energy varied from 0.33 to 0.47 kW·h·m-3respectively,depending on the specific layout.With the selectivity around 29 and CO2permeance around 2.5×10-2mol·(m2·s·MPa)-1(74 GPU),the two-stage membrane process with a cycle gas operated at 2.6 MPa offered the best performance,with a specific separation energy of 0.33–0.38 kW·h·m-3(equal to 1.18–1.37 MJ·m-3),which was slightly higher than 0.29–0.30 kW·h·m-3(equal to 1.04–1.08 MJ·m-3)reported by Deng and H?gg[44],due to the low membrane selectivity in Valenti's study.The simplicity of the two-stage membrane process was the winning factor for industrial application,and future work should be focused on applying a hollow-fiber membrane module which could effectively reduce the expense of biogas upgrading.

The impact of CO2/CH4selectivity and membrane area on separation performance was conducted by Makaruk and coworkers[45].A biogas upgrading case with a feed gas flow rate of 1000 m3·h-1,feed gas methane concentration of 60%and product gas methane concentration of 98%was investigated.The energy consumption for upgrading biogas to natural gas substitute was around 0.3 kW·h·m-3(equal to 1.08 MJ·m-3).The result showed that the increase of membrane area contributed to relatively higher product recovery.However,if a very high recovery close to 100%was achieved,considerably more membrane areas should be consumed.When a lower CO2/CH4selectivity membrane was employed, more compression power would be consumed to realize the same methane recovery.With a CO2/CH4selectivity higher than 50,the energy consumption could only slightly decrease.In this case,the membrane-based gas separation processes could provide enough flexibility to keep a high product recovery during the biogas upgrading process.

Scholz et al.[46]applied a structural optimization approach for membrane-based biogas upgrading processes to minimize both required membrane area and pressure,and presented the most profitable membrane process.General Algebraic Modeling System(GAMS)was implemented to optimize the selectivity and permeance in the membrane process with fixed product purity and recovery.The two-stage membrane process was proven to be the optimal design,as shown in Fig.8.The CO2selectivity and permeance were determined by the Robeson upper bound[6],with CH4recovery fixed as 99.5%and CH4purity fixed as 96%.When each stage employed the same selectivity membrane,the optimal membrane permselectivity for the two-stage membrane process was a CO2/CH4selectivity of 123 combined with a CO2permeance of 555 GPU.When each stage employed different selectivity membranes,the selectivity should be optimized individually.For the optimal condition,a membrane with high CO2/CH4selectivity of 147 and CO2permeance of 349 GPU was applied for the first stage,while a membrane with a moderate CO2/CH4selectivity of 98 and CO2permeance of 1024 GPU was applied for the second stage.Due to the application of the Robeson upper bound,the required membrane area was significantly smaller than the commercial membranes,which only had a permeance of 60 GPU.As such,the membrane area consumption was only 1.95 m2·h·m-3CH4,operated at the feed pressure shown in Fig.8.The specific energy demand was 0.161 kWh·m-3(STP)with respect to the raw gas flow rate(equal to 0.966 MJ·m-3CH4).In sum,the two-stage membrane process with membranes applied to the Robeson upper bound provided the best performance.

In conclusion,product purity requirement markedly determined the optimal CO2/CH4selectivity,CO2permeance and capture cost.For instance,in order to acquire higher product purity,a higher optimal CO2/CH4selectivity,lower CO2permeance and higher capture cost were needed.This reflected that typical membrane-based gas processes had to make a strong effort in acquiring high purity product,and it is relatively expensive to achieve ultrahigh purity product by membranebased technology in the biogas upgrading process.

In the study conducted by Kim et al.[47],hydrogen sulfide and carbon dioxide were removed from biogas using the membrane separation process and the purified methane was used as the fuel of fuel cell.The requirement of product in Kim's work was CH4purity above 99%and H2S concentration lower than 5×10-6,combined with CH4recovery above 90%.The membrane CO2permeance was around 75 GPU,and membrane CO2/CH4selectivity was around 15.Numerical solution obtained from MATLAB was employed to determine the relationship between CH4recovery and membrane area consumption of each stage in a multi-stage separation process with recycle stream.The three-stage membrane process with recycle stream was optimized to realize a CH4recovery of 91%and CH4purity of 99.7%with a membrane area of 454.3 m2(equal to 65.4 m2·h·m-3CH4).Increasing operating pressure from 0.388 MPa to 0.679 MPa would lead to the reduction of membrane area and also the decrease of CH4recovery.As such,Kim concluded that the key optimization variables were the operating pressure and membrane area for the removal of H2S and CO2from biogas.

Typically,membrane-based gas separation processes are modeled in steady state.However,in practical,it is inevitable to encounter the fluctuation of feed gas composition and flow rate during the operation.In these cases,product composition should be kept as constant through a dynamic process control system during the separation process.To resolve the above problem,Scholz et al.[48]presented efficient process control schemes to maintain the product purity and CH4recovery as constant simultaneously during the biogas upgrading process.Two novel process control schemes programmed in Aspen Custom Modeler and simulated in Aspen Plus Dynamics were designed to realize the above target,even if significant changes in the feed gas conditions were still encountered.The membrane CO2permeance was 60 GPU,and the CO2/CH4selectivity was 60,with the operating temperature of 298 K and feed pressure of 1.6 MPa.

Fig.8.Optimal process design for a gas permeation material with Robeson upper bound characteristics[46].1 bar=0.1 MPa.

The process control schemes depicted in Fig.9 were able to maintain product CH4concentration as 96%.The conditions of feed gas were measured online,so as to provide the necessary data for a model predictive controller(MPC)that controlled product purity.The PID controller adjusted the pressure according to the product purity measured.If the feed condition deviated from the design point,lower CH4recoveries or increased cycle gas flow rate was expected.This process could be applied to the start and shut down processes in gas permeation processes,and could also be used to investigate the impact of material degradation on the gas process during the operation.

The hybrid processes which combined the membrane process with other traditional gas separation processes such as pressurized water scrubbing,amine absorption and cryogenic separation were investigated by Scholz et al.[49]for biogas upgrading.The investigation of hybrid processes was simulated by Aspen Plus and the cost estimation was based on Guthrie's method.The combination of membrane technology and existing separation techniques was proven to have better performance than merely using the traditional cryogenic separation process or amine absorption process,and both lower cost and higher CH4recovery could be obtained for the hybrid processes.As such,when the membrane process was incorporated into the established processes,the energy consumption and capture cost could be further reduced.

Fig.9.Two advanced control schemes which combine a conventional controller with a model predictive controller.(a)Model predictive controller(MPC)sets the permeate pressure of the first stage.(b)Model predictive controller(MPC)sets the permeate pressure of the second stage permeate[48].

In most situations,a constant membrane permeance was often assumed for the sake of simplicity during the simulation.In the study conducted by Bounaceur et al.[50],the impact of pressure on the permeance of different gases was investigated systematically based on the solution–diffusion model.The result showed that constant permeance could lead to significant errors about membrane area consumption in the CO2/CH4separation process.The author recommended that if the CO2concentration difference between feed gas and residue gas was higher than 1%,the model assuming constant of membrane permeance was not accurate enough for predicting the membrane separation performance.Even though using average permeance could provide a reasonable result,when the permeance for only one of the gases changed.If one of the gas permeances changed slightly during the separation process,an average permeability approach could provide a reasonable result.Therefore,the impact of variable permeance should be taken into account during the optimization of capital costs,operational costs and energy consumption,if a variable permeance behavior was observed in membrane module.

It is speculated that membrane-based separation processes will be frequently used as a practical biogas upgrading technology in the future.Both high CH4purity and high CH4recovery can be achieved by the twostage membrane process.The improvement of membrane CO2/CH4selectivity and permeance further could reduce the application cost of the membrane-based gas separation process.There are still some conundrums needed to be resolved or investigated in the future as below.

1.There is a lack of systematic investigation on how operating conditions and membrane materials affect the operating cost with fixed product purity and recovery.Therefore,systematic investigations of biogas upgrading involving both fixed separation target and optimal operating cost are urgently needed.

2.Further improvement of membrane selectivity or permeance might not bring about additional benefits for biogas upgrading.Systematical investigations of the impact of membrane selectivity,membrane permeance and operating pressure on capital cost,operational cost and energy consumption are required for researchers to understand the required membrane permselectivity with optimal operating cost.

3.The two-stage membrane process should be investigated with more detail,since it is the simplest upgrading process to satisfy the requirement of CH4concentration above 97%and CH4recovery above 98%,which is appealing for bio-methane production.

4.If the permeance of gases changes significantly during the membrane separation process,membrane permeance calculated in the membrane process should be adjusted according to the specific operating condition.

5.The high selectivity membrane is of great importance to the realization of the biogas upgrade process.The PIM-1/Matrimid membrane[51]possesses a CO2/CH4selectivity of 34.3 and CO2permeance of 243.2 GPU,which is recommended for the potential application in the biogas upgrade process.

2.3.Natural gas carbon dioxide removal

Natural gas(NG)is one of the main sources of energy supply.Raw natural gas[52]typically contains 75%–90%CH4,carbon dioxide,nitrogen and hydrogen sulfide.Actually,the CO2concentration and H2S concentration vary in different geographic locations,while methane is always the major component.Since CO2could reduce the heating value of natural gas,freeze to form dry ice,which would block pipeline and damage pumps,the U.S.rolled out increasingly stringent mandatory measures to stipulate the US pipeline specifications of CO2≤2%and H2S ≤4×10-6for CO2removal from natural gas[4].Since 1980s,the CO2/CH4membrane has been commercialized for natural gas carbon dioxide removal processes[52].The membrane separation process provides significant advantages for the offshore applications due to the small footprint and environmentally friendly operations.Many researchers[53,54]proved the feasibility of membrane processes for natural gas purification,for the driving force for membrane separation could be directly obtained by the high wellhead gas pressure of raw natural gas.

In order to achieve the US pipeline specifications(≤2 mol%CO2and≤4×10-6H2S)of natural gas,single-stage and two-stage processes are two alternative kinds of processes for the separation aim,and the CO2-selective membrane and H2S-selective membrane are investigated to find out the most economical process design.In general,natural gas was simulated by a ternary gas mixture containing 0–40%CO2and 0–10%H2S balanced with CH4,and feed pressure was 800 psi(5.52 MPa)and temperature was 35°C,and permeate pressure was 20 psi(0.138 MPa).The results offered by Hao et al.[55]showed that natural gas containing large amount of H2S(≤10 mol%)and small amount of CO2could be treated economically with H2S-selective membranes with a CH4permeance of 20 GPU,CO2/CH4selectivity of 16 and H2S/CH4selectivity of 75 in the single membrane process.While natural gas containing large amount of CO2(≤40 mol%)and small amount of H2S(≤8×10-6)could be purified economically with CO2-selective membranes with a CH4permeance of 1 GPU,CO2/CH4selectivity of 60 and H2S/CH4selectivity of 15 in the single membrane process.When both the amount of CO2(up to 40%)and the amount of H2S(up to 10%)were large enough,the two-stage membrane process with the H2S-selective membrane in the first stage and the CO2-selective membrane in the second stage was the most economical process configuration.It was obvious that the membrane area requirement acted as a function of both the CO2and H2S concentration in the feed gas.For example,if the feed gas contained 0–40 mol%CO2and 1 mol%H2S,total membrane area requirement varied from 2373.7 m2to 3681.8 m2,and if the feed gas contained 1×10-6–10 mol%H2S and 20 mol%CO2,total membrane area requirement varied from 14477.1 m2to 128915.1 m2,and then the processing cost also changed with the composition of feed gas.Therefore,by applying different selective membranes in each stage,the two-stage membrane process with recycle stream could bring about high CH4purity,high recovery and low upgrading cost.

Except for the commonly used two-stage membrane process and two-step process,many other complex processes also have been investigated.Further studies were conducted by Hao and coworkers[56],which investigated two-stage and three-stage membrane processes,with the same membrane permselectivity as reported in Hao's previous work[55].A sensitivity analysis was made to find the optimal process design with the best upgrading cost.The two-stage membrane process with the H2S-selective membrane in the first stage and either CO2or H2S-selective membrane in the second stage was found to provide the optimal upgrading cost.Under the conditions assumed in Hao's study,the three-stage membrane process was found to be less economically competitive than the two-stage membrane process.As a result,the two-stage membrane process with different kinds of membrane applied to each stage was proven to be a better process than three stage membrane processes,when applied to upgrade the low quality natural gas containing large amount of H2S and CO2.

In the work of Yang et al.[13],natural gas containing 10%CO2and 90%CH4was purified with single-stage and two-stage membrane processes shown in Fig.10.The results showed that the increase of feed pressure or the decrease of permeate pressure could effectively reduce the membrane area consumption and improve the CH4recovery.With membrane selectivity above 50,the separation target of CH4recovery＞95%and CH4purity ＞98%could be achieved by the single-stage membrane process under the feed pressure of 5 MPa and the permeate side pressure of 0.12 MPa.The total membrane area requirement should be tens of thousands of square meters with a feed flow rate of 11.47 m3(STP)·day-1[13].For the two-stage membrane process,a low membrane selectivity of 20 was proven to be high enough to reach the same separation target,since the recycle gas managed to improve methane recovery.Hence,for the sake of reducing the membrane area consumption,moderate selectivity(i.e.,slightly lower than 100)and relatively high permeance were required.The energy compression for the two-stage membrane process was estimated to be 107.5 kJ·m-3,which was much lower than that of the amine absorption process,even considering the conversion factor between heat and electricity of 3–4.

Fig.10.Schematic diagram of single-stage and two-stage membrane process for natural gas carbon dioxide removal.

Most of the previous studies did not consider non-ideal effects such as pressure and temperature,which might have an impact on membrane permeance.With the decrease of gas pressure,the Joule–Thomson(JT)effect takes its place and it could lead to temperature drop in the permeate gas,which could progressively influence the membrane permeance.Ahmad et al.[57]investigated the impact of the JT effect on membrane permeance for the separation of CO2from natural gas with a feed temperature of 50°C,feed pressure of 5 MPa and permeate pressure of 0.12 MPa.Pressure dependence of membrane permeance was compiled by Visual Basic(VB)program using the finite element method and implemented into an Aspen HYSYS user defined unit.The simulate results were validated by experimental data published,and the two exhibited great agreement.The comparison between the results of the ideal model and non-ideal model showed that the non-ideal model provided higher retentate gas CO2concentration,lower stage cut and lower methane loss than the ideal model.Feed gas with higher CO2concentration intensified the non-ideal effect which significantly could affect product purity,methane recovery,stage cut,compressor power consumption and natural gas processing cost.Thus,it is important to take the nonideal effect into consideration,when accurate separation performance and economics of gas separation system are expected.

Scholz et al.[14]studied the non-ideal effect in the membrane process with different feed temperatures and feed pressures,and investigated more details(such as considering concentration polarization,JT effect,pressure losses and real gas behavior)than the study of Ahmad and coworkers.The results showed that the non-ideal behavior caused by the JT effect during the separation of CH4and CO2should be taken into account when the feed pressure was above 1 MPa.The differences of the retentate gas CO2concentration between the ideal gas and nonideal model became more pronounced with feed pressure more than 3 MPa.Compared with non-ideal behavior caused by the JT effect,the impact of pressure loss and concentration polarization was negligible.Therefore,the JT effect should be taken into account for large-scale process design with feed pressure above 1 MPa.As such,it was expected that researches on natural gas carbon dioxide removal should address the JT effect in the separation process,if the operating pressure was equal to the wellhead gas pressure.

Khalilpour et al.[58]investigated the impact of feed gas quality,feed pressure,membrane area,selectivity and permeance on separation performance to better understand the design optima in the single-stage membrane process for natural gas carbon dioxide removal.The results showed that a medium selectivity of 60 combined with a medium permeance of(300–500)×10-10mol·m-2·s-1·Pa-1(90–150 GPU)at the feed temperature of 25°C was more advantageous in terms of process cost.High selectivity or permeance was shown not to be necessary for natural gas carbon dioxide removal.The optimal selectivity or permeance could be determined through techno-economic analysis.

Kwon et al.[59]investigated different membrane configurations via a new approach that visualized the economic performance for CO2removal from natural gas through a case study using a hollow fiber membrane with the feed pressure of 8 MPa,permeate pressure of 0.4 MPa and flow rate of 32.8 m3·s-1.The membrane maintained a CO2permeance of 40 GPU,CO2/CH4selectivity of 36,C2H6/CH4selectivity of 0.27 and C3H8/CH4selectivity of 0.054.Response surface methodology(RSM)was adopted in Kwon's work to find the optimal operating parameters with minimum economic cost.Based on this method,the results were presented in a 3-D graph in which the effects of different parameters could be compared simultaneously to identify the optimal condition based on the same separation criteria.The two-stage membrane system was found to exhibit low membrane area consumption to match the pipeline specification(CO2concentration below 2%),and it was taken as the most profitable configuration in Kwon's study.The graphs of response surface could be used to find out the most appropriate configuration with the minimal investment cost,showing large numbers of design parameters simultaneously.Therefore,RSM could be a very useful optimization method in process design.

According to the discussions above,there are some suggestions for the natural gas carbon dioxide removal processes below.

1.It is the high pressure that makes the gases behave non-ideally and the Joule–Thomson effect rather pronounced,which could reduce the membrane temperature and permeance,then progressively causing the increase of membrane area consumption.As a result,for the natural gas membrane separation process,the Joule–Thomson effect should be taken into consideration when the membrane process is operated at a high pressure above 10 bar.

2.If the H2S concentration in raw natural gas is too high to be neglected,the membrane process with the H2S-selective membrane and CO2-selective membrane in different stages should be discussed with more details.

3.The two-stage membrane process was found to be more economically competitive than the three-stage membrane process.Over time,the optimization of the two-stage membrane process will increase the competitiveness of membrane processes for natural gas carbon dioxide removal.

4.When membrane selectivity is above 20,the separation target of CH4recovery ＞95%and CH4purity ＞98%could be achieved by the twostage membrane process under the feed pressure of 5 MPa.Besides,for the sake of reducing the membrane area consumption further,higher selectivity(i.e.,20–100)and relatively high permeance are required.

The typical designs for membrane-based gas separation processes include single-stage,two-stage/two-step,and even three-stage membrane processes.It is worth mentioning that when the three-stage process is applied,the immense complication and high energy consumption are incurred for the separation process.As for the single stage/two-step membrane processes,it is hard to achieve both high product purity and high recovery,based on the membrane properties(selectivity and recovery)reported.Therefore,it is the two-stage membrane process,as this paper mainly reviewed,that is well recommended to achieve both high purity and high recovery,with the most cost and energy saving way.

3.Conclusions

This paper systematically reviews the membrane processes for postcombustion CO2capture,biogas upgrading and natural gas carbon dioxide removal.Depending on the specific applications(such as feed pressure, compositions of feed gas) of the CO2membrane, the corresponding requirements on membrane materials and operating conditions have significant differences.The single-stage membrane process is not an effective option if the feed gas CO2concentration is too low or the product purity requirement is too high.As a result,two-stage membrane processes as well as multi-stage membrane processes can be regarded as an effective option for CO2separation.Validated by genetic algorithm and systematic analysis,two-stage membrane processes are proven to be more attractive than multi-stage (＞2)membrane processes for the realization of high product purity and recovery,when low energy consumption and operating cost are desired.As such,being an immature technology,the two-stage membrane system,which is on the path from laboratory to industrial scale,should still be investigated deeply with more details in the future.Therefore,the pilot scale process plays a really critical role,and it is urgent for the researchers to provide more data in a pilot scale test to accelerate the industrialization of the membrane-based CO2separation process.