CHAPTER 12 |
 |
Simulation of Hydrogen Purification by Pressure-Swing Adsorption for Application in Fuel Cells
Qinglin Huang and Mladen Eić
Department of Chemical Engineering, University of New Brunswick, Fredericton, New Brunswick, Canada
Contents
4.2. Parametric Studies of PSA System
1. INTRODUCTION
Increasing public pressure to reduce greenhouse gas emissions and limited resources in supplying fossil fuels have led to demands for diversifying energy resources as a priority. Hydrogen is a clean and sustainable energy resource that only emits water and heat resulting from the combustion. One of the important hydrogen applications is its use in proton-exchange membrane fuel cells (PEMFC), which have been considered an attractive technology for transportation, commercial building, home, and small-scale power generation. Fuel cell converts the electrochemical energy stored in hydrogen into electricity and requires high hydrogen purity, i.e., greater than 99.99%. Impurities were reported to have negative impact on fuel-cell performance and durability, and the fuel composition must therefore be carefully controlled [1, 2]. Among these impurities, CO is one of the most critical ones affecting fuel-cell operation. CO poisons the catalyst by occupying the active sites and substantially reducing the number of sites for hydrogen adsorption and oxidation. The concentration of CO in fuel-cell feed is recommended to be below 0.2, 0.5, or 10 ppm, depending on standards used [2–4]. Other impurities, such as CO2 and CH4, have less negative impact on fuel-cell performance compared to CO. Concentrations of CO2 and CH4 are proposed to be lower than 2 and 100 ppm, respectively [1, 2].
High-purity hydrogen (up to 99.999%) can be obtained by the noncryogenic pressure-swing adsorption (PSA) process from the steam methane reforming (SMR) process, which produces approximately 95% of hydrogen in the entire hydrogen industry. The typical feed gas compositions from SMR to PSA process contain 70–80% H2, 15–25% CO2, 3–6% CH4, 1–3% CO, and trace N2 [5]. A number of experimental and theoretical studies have been carried out to investigate hydrogen separation from SMR off-gas, oven coke gas, or refinery fuel gas by PSA process using single/layered-bed configuration [6–9]. However, only few studies have been focused on controlling hydrogen impurity for fuel-cell application. Moreover, the optimization work is necessary for PSA system due to the complicated nature of dynamic cyclic process, and the fact that there are a large number of design parameters, such as step times, pressure, temperature, gas velocity, and bed dimensions, which can affect PSA separation performance. For that reason, it is not always feasible to carry out a large number of PSA experiments to determine the optimum conditions. So far, there have been several studies investigating the optimization of single-bed PSA process using gPROMS optimizer [10–13]. However, none of them studied optimization of PSA process with layered adsorbents in bed and using a feed that has more than two constituent components.
The objective of this work was to theoretically investigate the feasibility of hydrogen production from SMR off-gas by layered-bed PSA to meet the purity requirements for fuel-cell application. Activated carbon and zeolite 5A were used as adsorbents, whereas concentrations of CO, CO2, and CH4 were targeted to be below 10, 100 and 100 ppm, respectively. Optimization of single-bed PSA process with layered adsorbents to separate hydrogen from quaternary mixture of CO2, CO, CH4, and H2 using gPROMS optimizer tool was also carried out for comparison.
2. PSA MODEL AND SOLUTION
2.1. Model Assumptions
Mathematical model was applied to describe two-layered-bed PSA system using six-step cycle (Fig. 12.1). The following assumptions were used in the model development:
1. The ideal gas law applies.
2. The pressure drop through the bed during adsorption and purge steps is described by Ergun’s Equation. For other steps, the total pressure in the bed is assumed to vary linearly with time.
3. The flow pattern is described by an axially dispersed plug flow model.
4. The multicomponent adsorption equilibrium is represented by the extended Langmuir isotherm.
5. There are no radial variations in temperature, pressure, concentration, or velocity.
6. Transport and physical properties are temperature independent.
7. The adsorption rate is approximated by a linear driving force (LDF).
Figure 12.1 Six-step, two-layered-bed PSA cycle: 1, feed pressurization from intermediate to high pressure; 2, adsorption at high pressure; 3, pressure equalization between two beds from high to intermediate pressure; 4, depressurization from intermediate to low pressure; 5, purge with product from the other bed at low pressure; 6, pressure equalization between two beds from low to intermediate pressure.
2.2. Model Equations
The component and overall mass balances for the system are represented by Eqs. (1) and (2), respectively:
Equation (2) determines the change of gas interstitial velocity over the length of the bed. In Eqs. (1) and (2), u is gas interstitial velocity; ci is gas phase concentration of component i; C is total concentration in gas phase (= P/RgT); Rg is ideal gas constant, T is temperature; z is axial co-ordinate; qi is adsorbed phase concentration of component i; ρp is particle density; εbed is bed voidage; DL is the axial dispersion coefficient, which can be estimated by the following equation [14]:
where dp is the diameter of adsorbent particle; Dm is the molecular diffusivity that can be calculated from the well-known Chapman–Enskog or similar equation.
Heat effects are described by the energy balance equation:
where Cpg and Cps are heat capacities of gas and solid particles, respectively; ρbed and ρgas are gas densities of bed and bulk, respectively; h is heat transfer coefficient; KL is thermal dispersion coefficient; Rbed is column radius; Twall is column wall temperature; ΔHi is isosteric heat of adsorption, which is obtained from the Clausius–Clapeyron equation at the constant loading:
where Pi is partial pressure of component i. The adsorption rate is represented by the LDF model:
where ki is the mass-transfer coefficient; 
is the equilibrium loading of component i at time t, and can be determined by the extended Langmuir isotherm:
where bi is single-component equilibrium constant; qs is saturation capacity; b1,i and b2,i are temperature-independent parameters. The pressure drop across the bed during adsorption and purge steps is described by Ergun’s equation:
It should be noted that the layered-bed PSA system, investigated in this study, contains an initial section of activated carbon, followed by zeolite 5A in the remaining of the bed. Hence, in the PSA simulation model, the adsorbent physical properties, as well as the equilibrium parameters, mass-transfer and heat-transfer coefficients of the sorbates in the bed change at the transition location from one adsorbent to another.
All boundary conditions for the cycle steps with respect to the differential variables within the model, such as gas concentration, velocity, and temperature, are summarized in Table 12.1 [13]. The performance variables including purity, impurity, and recovery are defined as
Table 12.1 Boundary conditions of six-step PSA process
where tA, tP, and tPu are adsorption (second step in Fig. 12.1), pressurization (first step in Fig. 12.1), and purging (regeneration; fifth step in Fig. 12.1) times, respectively.
2.3. Solution Methodology
The partial differential equations are discretized into n discretization points using the centered finite difference method. It may be noted that the number of discretization points is important for the accuracy of the numerical solution. The optimum numbers were determined by increasing discretization points until further change did not affect the solutions any more. The coupled set of ordinary differential and algebraic equations are solved by DASOLV that is based on variable time step or variable-order backward differentiation formulae. The dynamic PSA simulation is carried out using gPROMS software licensed by Process Systems Enterprise Limited. The SRQPD solver that employs a sequential quadratic programming in gPROMS was applied to solve the nonlinear programming problem for PSA optimization.
3. EXPERIMENTAL
Single-component adsorption isotherms of CO2, CO, CH4, and H2 in activated carbon and zeolite 5A, which were provided by Barnebey Sutcliffe Co. and Union Carbide, respectively, were measured at different temperatures using a custom-made constant volumetric apparatus. The experimental procedure was described elsewhere [15].
4. RESULTS AND DISCUSSION
4.1. Adsorption Equilibrium, Mass-Transfer Coefficients, and Multicomponent Column Dynamics in a Layered Bed
The experimental adsorption isotherm results together with Langmuir fittings are shown in Fig. 12.2. The model fittings show reasonable agreement with experimental data. The experimental results are comparable to those reported in the literature [16–18]. The optimized equilibrium parameter values obtained from nonlinear regression method of Langmuir model fittings are summarized in Table 12.2 (columns 3–5). Henry constant values at 303.15 K were obtained from the linear portion of experimental isotherm data. Evidently, activated carbon exhibited higher saturation adsorption capacities for CO2 and CH4 and lower saturation adsorption capacities for CO compared to zeolite 5A. Furthermore, the adsorption affinity is stronger for CO than CH4 in zeolite 5A. As can be seen from Fig. 12.2, zeolite 5A shows nearly rectangular isotherm for CO2, e.g., very high Henry constant value, indicating a difficulty to perform the desorption step at low pressures. Hence, CO2 had to be removed by activated carbon first to prevent it from breaking through into the subsequent layer of zeolite 5A.
Figure 12.2 Adsorption isotherms of CO2, CO, CH4, and H2 on activated carbon and zeolite 5A at three different temperatures (symbols: experiments; lines: Langmuir fit).
Table 12.2 Equilibrium and mass-transfer parameters for CO2, CO, H2, and CH4 in activated carbon and zeolite 5A samples
The overall mass-transfer coefficients (k) were obtained by fitting the single-component column dynamic breakthrough data measured at low sorbate concentration with helium as carrier gas [15]. The k values for CO2, CO, CH4, and H2 in activated carbon and zeolite 5A are included in Table 12.2 (last column).
Figure 12.3 shows the simulation of quaternary CO2, CO, CH4, and H2 breakthrough curves at different locations on the layered column.
Figure 12.3 Simulated concentration breakthrough curves for CO2, CO, CH4, and H2 at 298.15 K, and z = 0.4, 0.6, 0.65, 0.7, 0.8, and 1.0 of a layered column: column was initially saturated with H2 at 12 atm and then fed with 23%CO2/2%CO/3%CH4/72%H2 gas mixture.
The adsorbent properties and other relevant information of layered column are listed in Table 12.3. The equilibrium and mass-transfer parameters in Table 12.2 were used in the simulation. It is clear from Fig. 12.3 that the gas concentrations at different bed positions asymptotically approach the feed mixture at 23%CO2/2%CO/3%CH4/72%H2, after sufficiently long time. The breakthrough times at the column exit (z = 1.0) are tCH4 < tCO < tCO2 which was also reported by Park et al. [17]. As can also be observed from the same figure, CO mass transfer wave moves faster than CH4 in the activated carbon layer (z ≤ 0.6). After breaking through into the zeolite layer (0.6 1 z ≤ 1.0), CH4 wave velocity increases relative to CO. CH4 wave caught up the CO front at z = 0.7 and surpassed it as z 2 0.7, e.g., z = 0.8 or 1.0. This phenomenon could be explained from the definition of concentration wave velocity given by [19]:
Table 12.3 Properties of adsorbents and layered column information
where dq/dc is Henry’s constant in the linear region of isotherm. Henry’s constants in Table 12.2 are in the sequence CH4 2 CO and CO 2 CH4 for activated carbon and zeolite 5A, respectively. This demonstrated that velocity of CO wave was greater than CH4 in the activated carbon layer, whereas the opposite was true in the zeolite layer.
From Fig. 12.3, concentration roll-ups for both CO and CH4 can also be observed. The concentration wave of CH4 exhibits double roll-ups at z = 1.0 due to the breakthrough of CO followed by CO2. Minor transitions can also be observed from the hydrogen desorption profiles, relating to the breakthroughs of other components. For multicomponent adsorption, the relatively weakly adsorbed species are always displaced by the more strongly adsorbed components. The roll-ups of one component generally occur when the other components break through the adsorbent layer with a steep wave front [17].
Figure 12.3 shows that CO and CH4 concentration profiles become concave after the roll-ups due to the adsorption heat effects. The temperature profiles (Fig. 12.4) in the layered bed can account for the observed phenomena if the concave section of concentration curves corresponds to the temperature peak. The adsorbed amount in the adsorbents was reduced due to the substantial increase of temperature, rendering the reduction of gas-phase concentration in the bed. Figure 12.4 shows that there are two separate temperature peaks in the zeolite layer. The first one is due to the adsorption of CO and CH4, whereas the second one is due to adsorption of CO2. Because CO2 has higher heat of adsorption in zeolite 5A than in activated carbon, the related temperature peak in zeolite layer is correspondingly larger.
Figure 12.4 Temperature profiles at z = 0.2, 0.4, 0.65, and 1.0 of a layered column. For simulation conditions, see Fig. 12.3.
4.2. Parametric Studies of PSA System
The parametric effects on PSA separation performance were theoretically investigated in this section. The results were collected at the 150th cycle of PSA simulation, which approximately reached a cyclic steady state (CSS).
4.2.1. Effects of Interstitial Velocity on the Adsorption Step, Purge Time, and Feed Temperature
In the commercial PSA system, the interstitial velocity during adsorption step is one of the most important operating parameters and, therefore, has to be carefully selected. Figure 12.5A shows the simulation results related to the effects of the interstitial velocity on the separation performance of the PSA system. The product impurity was dominated by CO and then CH4 as the velocity increased. To obtain fuel-cell-grade hydrogen, the velocity should be maintained below 5 cm/s, under the operating conditions studied here. Hydrogen purity decreased sharply with velocity greater than 5 cm/s, whereas the recovery gradually increased with the increase of velocity. When the velocity was increased, the breakthrough times decreased; consequently, within the constant cycle time, this caused a decrease in the product purity. The decrease in purity is dependent on the extent of impurity breakthrough. For the lower velocity runs, e.g., u 1 5 cm/s, the impurity breakthrough was insignificant; therefore, the purity dropped only slightly. On the other hand, at higher velocities, the impurity breakthrough became more significant, causing a rapid drop in purity.
The effect of the purging time on the performance of the PSA process is illustrated in Fig. 12.5B. CO was dominant among impurities and could be controlled to be less than 10 ppm (mole ratio of 1.0 × 10−5 marked in the figure), when the purging time was greater than 177 s. The increase in purging time corresponds to more thorough cleaning of the adsorption bed, rendering higher purity, and lower product recovery.
The effects of the temperature on the performance of the PSA process are shown in Fig. 12.5C. Apparently, high temperatures are generally unfavorable for impurity removal. The product recovery, as well as CO and CO2 impurity, increased with the increase of adsorption temperature (with CO dominating as the product impurity). The adsorption temperature was required to be lower than 299 K for controlling CO impurity to be lower than 10 ppm. CH4 impurity was very low (mole ratio approximately 1.0 × 10−9) when the adsorption temperature was below 318 K; thus, CH4 could be considered completely removed at these temperatures. However, it increased at the adsorption temperatures above 318 K. With an increase in the temperature, the adsorption capacity of the adsorbents decreased, leading to shorter breakthrough time and hence lower product purity at the constant cycle time.
Figure 12.5 Effects of (A) adsorption velocity, (B) regeneration time, and (C) adsorption temperature on PSA separation performance. Basic PSA operating conditions, i.e., adsorbent ratio, 0.6; feed composition, 23%CO2/2%CO/3%CH4/72%H2; feed temperature, 298 K; feed pressure, 12 atm; purge pressure, 1 atm; feed pressurization time, 20 s; adsorption time, 180 s; first pressure-equalization time, 20 s; depressurization time, 20 s; purge time, 180 s; second pressure-equalization time: 20 s; adsorption and purge velocity, 5 cm/s, were used in the simulation. In the parametric studies, the investigated parameter was varied, while all others were kept constant.
4.2.2. Effects of Adsorption Pressure and Adsorbent Ratio
The effects of adsorbent ratio and pressure on the PSA separation performance were investigated for constant feed and regeneration flow rates, as well as same step times. The results are shown in Fig. 12.6A–E. Here, the adsorbent ratio was defined as a ratio of activated carbon layer length to the packed bed length.
Figure 12.6A–C shows that high adsorbent ratio favors CO2 and CH4 removal, whereas low ratio is favorable for CO removal at all pressures used in this study. This is expected because the adsorption capacities of CH4 and CO2 are higher, whereas the capacity of CO is lower in the activated carbon compared to the zeolite 5A. The high interstitial velocity in the bed, which corresponds to the low operating pressure, e.g., P = 8 atm, results in a complete or partial breakthrough of impurity, thus reducing the product purity that becomes too low to be used for fuel cell. However, at higher pressure (P = 12, 16 atm), CH4 impurity is always less than 15 ppm (mole ratio, 1.5 × 10−5), and also, there is a range of adsorbent ratios (0.35–0.92 and 0.44–0.61 at P = 16 and 12 atm, respectively) that allows control of CO2 and CO to the contents of less than 100 and 10 ppm, respectively. Such range is even broader at the higher adsorption pressures. There is also an optimum range of adsorbent ratios (0.39–0.90 and 0.50–0.80 at P = 16 and 12 atm, respectively, shown in Fig. 12.6B) for which minimum CO2 impurity can be obtained, i.e., CO2 can be effectively prevented from breaking through into the zeolite 5A layer.
As shown in Fig. 12.6D and E, hydrogen purity is increased by increasing pressure, while the product recovery is decreased. Such effect is similar to that of interstitial velocity as shown in Fig. 12.5A. With constant feed flow rate, the interstitial velocity is lower at the elevated adsorption pressure, leading to higher hydrogen purity and lower recovery. From Fig. 12.6D (and the inset), it can also be observed that for each adsorption pressure, there is one corresponding peak related to the hydrogen purity. Nearly, 100% H2 could be obtained from layered-bed PSA within adsorbent ratio 0.39–0.7 at P = 16 atm, whereas at P = 12 or 8 atm, there is a unique adsorbent ratio (0.48 and 0.8 at P = 12 and 8 atm, respectively) that generates the highest product purity. In the adsorption step of PSA operation, there is a critical impurity concentration wave velocity (see Eq. (13)), ωcritical, where complete impurity removal is attained as ωc ≤ ωcritical, while impurity is present in the product as ωc 2 ωcritical. At the higher adsorption pressures, e.g., P = 16 atm, the interstitial velocity is lower and the resulting ωc, for a range of adsorbent ratios, may be even lower than ωcritical, indicating the complete impurity removal. On the other hand, it is impossible to obtain pure product at the lower adsorption pressures (e.g., P = 8 and 12 atm corresponding to the higher interstitial velocities) because ωc for all adsorbent ratios is higher than ωcritical. However, there may exist a unique adsorbent ratio that can generate the maximum product purity less than 100%. As can be seen from Fig. 12.6D, due to the breakthrough of impurity, the product purity dropped sharply when adsorbent ratios were lower than 0.38, 0.43, and 0.7 at P = 16, 12, and 8 atm, respectively. Furthermore, as can be seen from Fig. 12.6E, adsorbent ratio of 0.1 produced the highest hydrogen recovery for each adsorption pressure. These results show that a layered-bed PSA performs better than either single-layered activated carbon or zeolite 5A bed, with respect to purity or recovery.
Figure 12.6 Effects of adsorption pressure and adsorbent ratio on PSA separation performance: impurities of (A) CO, (B) CO2, and (C) CH4, and (D) H2 purity in the product, and (E) H2 recovery. For PSA simulation conditions, see Fig. 12.5.
4.2.3. Effects of Feed Composition and Adsorbent Ratio
Because both activated carbon and zeolite 5A exhibited the highest adsorption capacities for CO2 among the four sorbates, three gas feeds with different compositions of CO2, i.e., 16%CO2/2%CO/3%CH4/79%H2, 23%CO2/2%CO/3%CH4/72%H2, and 30%CO2/2%CO/3%CH4/65%H2 designated as feed A, B, and C, respectively, were used to study their effects on PSA separation performance. PSA operating conditions were kept the same in the simulation. The simulation results are illustrated in Fig. 12.7A–E.
As shown in Fig. 12.7A–C, the entire adsorbent ratio range could generate CH4 impurity less than 15 ppm for all CO2 feed concentrations, whereas the adsorbent ratio range that could produce less than 100 ppm CO2 and 10 ppm CO is more restricted at the higher CO2 feed concentrations. The ratio ranges for production of fuel-cell-grade hydrogen are 0.23–0.50, 0.44–0.61, and 0.63–0.68 for CO2 feed concentration of 16, 23, and 30%, respectively. The increase in CO2 feed concentration requires an increase of the activated carbon layer, i.e., increase the adsorbent ratio for CO2 removal. However, at the same time, this would increase CO impurity in the product. Therefore, for each feed concentration, the adsorbent ratio range of layered PSA should be carefully set for producing fuel-cell-grade hydrogen.
Similarly, as stated in the above section, an optimum adsorbent ratio exists corresponding to the highest hydrogen purity for each CO2 concentration in the feed (see the inset in Fig. 12.7D). Hydrogen purity was found to be strongly dependent on CO2 feed concentration and adsorbent ratio. It decreased substantially with an increase in CO2 feed concentration, e.g., when the adsorbent ratio was below 0.26 for all CO2 feed concentrations, and 0.26–0.4 for feeds B and C. This is attributed to the increased extent of CO2 and CH4 breakthrough into the zeolite layer, which occurs at the higher CO2 feed concentration and lower adsorbent ratio. However, hydrogen purities, in the case where the main impurity was CO, were in the following sequence feed C 1 feed A 1 feed B and feed A 1 feed B 1 feed C as the adsorbent ratio was set between 0.5 and 0.7 and higher than 0.7, respectively. As adsorbent ratio increased, activated carbon layer was more and more efficient in removing most CO2 and CH4; thus, the remaining gases entering into the zeolite layer comprises mainly CO and H2. The lower CO2 concentration in the feed exhibited higher interstitial velocity, due to the higher H2 concentration. At the same time, the CO concentration into the zeolite layer was smaller compared with the case of the higher CO2 in the feed. However, the high interstitial velocity dominated the column dynamics in the zeolite layer, leading to the earlier breakthrough of CO. Therefore, feed A case even showed slightly higher CO impurity than feed B and C cases at high adsorbent ratios.
Figure 12.7 Effects of feed composition and adsorbent ratio on PSA separation performance: impurities of (A) CO, (B) CO2 and (C) CH4, and (D) H2 purity in the product, and (E) H2 recovery. For PSA simulation conditions, see Fig. 12.5.
The highest hydrogen recovery was observed, regardless of CO2 feed concentration, if the adsorbent ratio was kept at 0.1 (see Fig. 12.7E). If the adsorbent ratio was set to a higher value than 0.1, then hydrogen recovery decreased with that ratio. Furthermore, hydrogen recovery increased at higher CO2 feed concentration if the adsorbent ratio was kept at 0.1–0.7.
4.3. PSA Optimization
PSA processes are intrinsically dynamic, operating in a periodic fashion. The PSA separation performance depends on a number of design and operating parameters. gPROMS has been implemented in a number of PSA optimization studies [10–13]. Due to the limitation of gOPT – dynamic optimizer in gPROMS, e.g., limited optimization strategy and numerical solvers, only single-bed PSA process optimization was investigated in this study. For more complicated PSA system containing multiple beds and interactions among beds, e.g., pressure-equalization step, more specialized optimization strategy was proposed [20], which was not considered here. We attempted to apply gOPT to optimize a single-bed PSA system involving quaternary CO2, CO, CH4, and H2 mixture operated in four-step Skarstrom cycle: (1) countercurrent pressurization with hydrogen product to high pressure, (2) concurrent adsorption at high pressure, (3) countercurrent depressurization to atmospheric pressure, and (4) countercurrent regeneration at atmospheric pressure. Similar to the two-layered-bed PSA system studied above, the single bed was packed with activated carbon followed by zeolite 5A. The mathematical model for the single-bed PSA process, which is somewhat different in comparison to the two-layered-bed PSA described in the theory section, was detailed by Huang et al. [13].
For single-bed PSA with product pressurization process, the equation of product recovery is different:
In addition, the hydrogen product throughput is defined as
where tA, tP, tPu, and tcycle are adsorption, pressurization, purging, and cycle time of the four-step PSA, respectively.
The optimization of PSA model can be performed with respect to any number of performances such as product purity, product recovery, power, product throughput, etc. For performing PSA optimization, the PSA model was first simulated to CSS, which means that the mole fraction and solid concentration bed profiles were identical at the beginning and at the end of each cycle. The resulting bed profiles were fed to gOPT as an initial guess for the optimal CSS condition and were also discretized in the PSA model by method of lines. The chosen decision variables, such as cycle time, adsorption and regeneration velocities, adsorption pressure, etc. together with their constraints were added in gOPT. The optimization was then carried out by solving the objective function, as well as constraints on the variables. More details on optimization strategy were described in Ko et al. [11] and Huang et al. [13].
In this work, we investigated two optimization cases for separation of 23%CO2/2%CO/3%CH4/72%H2 based on maximum product purity and product throughput, and maximum product recovery. The constraints and objective functions of both cases are summarized in Table 12.4. Most constraints for both cases are the same except that case B includes impurity limits. It should also be noted that to find the desired optimal conditions, some constraints, such as adsorbent ratio, pressurization time, depressurization time, and regeneration pressure, were kept constant for the computation conversion. As discussed above, adsorbent ratio was an important factor affecting PSA separation performance. The value of adsorbent ratio was determined beforehand in the CSS simulation so that CO2 would not break into zeolite 5A layer, whereas CH4 and CO would not break for the whole bed.
Table 12.4 Optimization constraints and objective functions for two cases
The optimization results are listed in Table 12.5. Compared with case B, case A required much higher computational time due to the relatively more complicated objective function. Moreover, the adsorption time and regeneration time/velocity were increased for obtaining higher product throughput and product purity, respectively. However, the recovery for case A was lower than for case B. For case B, it was possible to obtain higher product recovery, as well as fuel-cell-grade hydrogen.
Table 12.5 Dynamic PSA optimization results for two cases
5. CONCLUSION
Single-component adsorption isotherms of CO2, CO, CH4, and H2 in activated carbon and zeolite 5A were experimentally measured and described by the Langmuir model. The two adsorbents exhibited various affinities and adsorption capacities for the selected probe species. Therefore, their respective effects on impurity removal from PSA process had to be explained and clarified.
A theoretical two-layered-bed PSA model has been developed for studying impurity removal from SMR off-gas for producing fuel-cell-grade hydrogen, where CO2, CO, and CH4 concentrations were required to be less than 100, 10, and 100 ppm, respectively. The column investigated in this study was packed with activated carbon followed by zeolite 5A. The effects of velocity during adsorption step, regeneration time, feed temperature, pressure and composition, and adsorbent ratio on hydrogen purity, impurity, and recovery were investigated methodologically. The results showed that the layered-bed PSA process was capable of providing higher purity than activated carbon or zeolite 5A single-layered PSA bed under the same operating conditions. A range of adsorbent ratios was also determined, which was shown to provide means for controlling the impurity concentration to satisfy requirements for fuel-cell application. Such range was dependent on operating conditions, such as feed pressure and composition.
The optimization routines were also carried out to examine the optimal conditions for the separation of 23%CO2/2%CO/3%CH4/72%H2 mixture using single-bed PSA process with layered adsorbents operated in Skarstrom cycle. Using the dynamic optimizer in gPROMS, the optimal operation conditions for PSA process based on different objective functions could also be obtained. gPROMS proved to be an effective tool for PSA simulation and simple optimization. However, optimization of more complicated PSA processes, including multiple beds and operating steps, requires further study.
ACKNOWLEDGMENTS
The authors gratefully acknowledge Dr Daeho Ko from GS Engineering and Construction Corp., Korea for stimulating discussions on gPROMS optimization for his contribution. The authors also acknowledge NSERC for the financial support.
References
[1] Cheng X, Shi Z, Glass N, Zhang L, Zhang J, Song D, Liu Z-S, Wang H, Shen J. J Power Sources 2007;165: 739–56.
[2] Besancon BM, Hasanov V, Imbault-Lastapis R, Benesch R, Barrio M, Molnvik MJ. Int J Hydrogen Energy 2009;34:2350–60.
[3] ISO. Hydrogen fuel – product specification – Part 2: proton exchange membrane fuel cell applications for road vehicles, ISO/TC 197 N 352, 2006.
[4] SAE. Information report on the development of a hydrogen quality guideline for fuel cell vehicles (SAE-J2719), 2005.
[5] Sircar S, Golden TC. Sep Sci Technol 2000;35:667–87.
[6] Malek A, Farooq S. AIChE J 1998;44:1985–92.
[7] Lee C-H, Yang J, Ahn H. AIChE J 1999;45:535–45.
[8] Lee J-J, Kim M-K, Lee D-G, Ahn H, Kim M-J, Lee C-H. AIChE J 2008;54:2054–64.
[9] Ribeiro AM, Grande CA, Lopes FVS, Loureiro JM, Rodrigues AE. Chem Eng Sci 2008;63:5258–73.
[10] Ko D, Siriwardane R, Biegler LT. Ind Eng Chem Res 2003;42:339–48.
[11] Ko D, Siriwardane R, Biegler LT. Ind Eng Chem Res 2005;44:8084–94.
[12] Knaebel SP, Ko D, Biegler LT. Adsorption 2005;11(Suppl):615–20.
[13] Huang Q, Malekian A, Eic M. Sep Purif Technol 2008;62:22–31.
[14] Ruthven DM. Principles of adsorption and adsorption processes. New York: John Wiley and Sons; 1984. p. 209.
[15] Malekian A. Adsorption characterization of novel nanoporous materials and optimization of bulk PSA process to produce H2 from plasma reactor. PhD thesis, University of New Brunswick, 2006.
[16] Yang J, Lee C-H. AIChE J 1998;44:1325–34.
[17] Park J-H, Kim J-N, Cho S-H, Kim J-D, Yang RT. Chem Eng Sci 1998;53:3951–63.
[18] Grande CA, F. Lopes VS, Ribeiro AM, Loureiro JM, Rodrigues AE. Sep Purif Technol 2008;48:1338–64.
[19] Farooq S, Huang Q, Karimi IA. Ind Eng Chem Res 2002;41:1098–106.
[20] Biegler LT, Jiang L, Fox VG. Sep Purif Rev 2004;33:1–39.