Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Steady-State Pressure-Swing

Adsorption (PSA) for Flowsheet


Simulation
Copyright (c) 2023 by Aspen Technology, Inc. All rights reserved.

Aspen Plus, Aspen HYSYS, and the aspen leaf logo are trademarks or registered trademarks of Aspen Technology,
Inc., Bedford, MA.

All other brand and product names are trademarks or registered trademarks of their respective companies.

Contents
Abstract..................................................................................................... 2
1 Background .......................................................................................... 3
2 Model Description .................................................................................. 3
3 Simulation Results ................................................................................. 7
4 Conclusion ............................................................................................ 8
References ............................................................................................... 10

Abstract
Pressure-swing adsorption (PSA) is common in applications such as air separation
and hydrogen purification from syngas, and is expanding to sustainability
applications, including H2 recovery after water electrolysis and CO2 capture. A
shortcut pressure-swing adsorption (PSA) example is presented in Aspen Custom
Modeler. The PSA is modeled as a steady-state, two-outlet separator described by
mass balance equations based on holdup and cycle time. This model could be further
exported and used in steady-state flowsheets in both Aspen HYSYS and Aspen Plus.

2
1 Background
Pressure-swing adsorption (PSA) is an established separation technology applied to
gas mixtures with moderate flow rates. While PSA units are already common in
applications such as air separation and hydrogen purification from syngas, the
technology also shows promise for ascendant sustainability applications, including H 2
recovery after water electrolysis and CO2 capture. In the plant, a PSA unit is a
complicated system comprising a minimum of two adsorbent-packed vessels (one
pressurizing and one de-pressurizing), and an automated system of valves to cycle
between states.

Aspen Adsorption provides an environment for detailed dynamic or cyclic steady-


state (CSS) simulation of PSA units. However, its results are not readily transferred
to a steady-state flowsheet for conceptual process design. With that in mind, Sees et
al. [1] proposed a simplified methodology to represent the average steady-state
performance of a PSA unit, using a system of steady-state separator blocks and a
series of calculations based on a limited set of top-level design parameters that the
user is likely to have at design time. In this example, the Sees et al. methodology is
recreated in Aspen Custom Modeler, resulting in a steady-state block that can be
exported to Aspen Plus.

2 Model Description
At the core of the Sees et al. methodology is a mass balance between a feed stream
having a steady flow rate n̄F , high-pressure and low-pressure products with steady-
state equivalent flow rates n̄HPP and n̄LPP. Conceptually, the configuration looks like
the flowsheet in Figure 1a (Figure 4a from Sees et al [1]). This system is a Skarstrom
PSA cycle with one modification, namely that initial pressurization n̄PR of the bed
coming online (before feed is introduced) is accomplished with the high-pressure
product stream instead of the feed stream. The high-pressure product is likewise
used to sweep (n̄SW) the offline bed after depressurization, and the internal high-
pressure and low-pressure retentate streams n̄HR and n̄LR represent the interstitial
and adsorbed gas left in the bed at the end of each step.

3
(a) (b)

Figure 1. Flowsheet representations of a Skarstrom PSA cycle with backfill. (a) From Sees et al [1]. (b)
Two-outlet splitter model in ACM.

For this example, the authors of the Sees et al. study provided AspenTech with
examples of this mass-balance calculation as (1) an Excel spreadsheet with circular
references, and (2) a sequential-modular flowsheet in Aspen Plus, comprising three
SEP blocks, two CALC blocks, and several internal recycle streams. The implicit mass
balance is easily handled with an equation-oriented model in ACM, as represented by
the two-outlet splitter shown in Figure 1b.

The present example is based on O2 purification from air. In the adsorption step, the
zeolite bed selectively adsorbs N2, concentrating O2 in the high-pressure product
(HPP). In the desorption step, the N2 is released from the bed as the low-pressure
product (LPP). A Langmuir isotherm gives the adsorbed holdup qads for each
component as a function of partial pressure:

,
= (1)
1+

Temperature dependence can be applied to the Langmuir parameters q and b with:

, = , + / (2)
1 1
= exp − (3)

From literature data, Sees et al. regressed the isotherm parameters in Table 1 for O 2
and N2 on Zeolite 5a. In practice, any isotherm formulation that gives qi as a function
of T, P, and the concentration of any interacting, co-adsorbed species could be
substituted for the Langmuir isotherm.

4
Table 1. Langmuir isotherm parameters for O2 and N2.

O2 N2

Saturation capacity (kmol/kg) 0.005062 0.005062

Xi capacity temperature dependence (kmol-K/kg) 0 0

Langmuir affinity constant at T0 (m3/kmol) 0.614 1.208

/ Reduced isoteric heat of adsorption (K) -1450 -1822

Reference temperature (°C) 45 45

In a typical use case, to predict the split of a given feed stream, the user must specify
the following:

 n̄F and {zf,i}, flow rate and composition of the feed stream

 {q }, isotherm for each component

 ρb and ε, bulk density and bulk void fraction for the adsorbent

 N, number of pressure vessels

 mads, mass of adsorbent per vessel

 PH, TH, PL, TL, temperature and pressure in the high- and low-pressure bed
states

 tcyc, estimate of total cycle time

Additionally, the following parameters can be specified to optimize or tune the


separation to match an existing unit:

 SW/F, the ratio of the sweep flow rate to the feed flow rate. A minimum sweep
rate can be estimated from the isotherm of the heavy key (N2 in this example)
and used in preliminary calculations.

 P/F, the ratio of the pressurization flow rate to the feed flow rate. Minimum
and maximum values for the pressurization flow rate can be estimated from
other parameters, with the maximum used in preliminary calculations.

 {ΦHP,i} and {ΦLP,i}, the so-called bed-state efficiencies for each component in
each step. These can be set to 1 for preliminary calculations.

Bulk concentrations are evaluated for three relevant states: the HP bed on fresh gas
feed (CHP), the LP bed after depressurization and sweep (CLP), and the HP bed during
pressurization (CPR). Given the scheme in Figure 1, where sweep and pressurization
are both supplied from the HPP stream, the formulas are:

5
,
, = (4)

z , P
C , = (5)
RT

,
, = (6)

The total gas holdups in each state X can be divided into the adsorbed material
remaining on the solid (Eq. 1) and interstitial material remaining in the void spaces
and

, = , ε/ρ (7)

, = , + , (8)

The flow and composition of the HP/LP retentate streams are then obtained from the
holdup divided by the cycle time. In the “one-efficiency model,” the bed-state
efficiency noted above is used to correct the relative composition as needed:

, =Φ , , / (9)

To complete the mass balance, the sweep and pressurization flow rates must be
specified. A minimum sweep rate can be computed from the Henry’s Law coefficient
for the heavy key kLP,N2.

, = , , / (10)

= , / (11)

Minimum and maximum pressurization rates are computed as follows:


ε
=( − ) / (12)
ρ

= + , − , / (13)

Let “raffinate” stand for the entire stream leaving the high-pressure bed, before it
splits into HPP, Sweep, and Pressurization:
= + − + (14)

, = , + , − , + , (15)

And the overall mass balance:


= − − (16)

= − (17)

6
3 Simulation Results
As noted above, the provided example is based on the literature comparison in Sees
et al, which was an air separation performed in a lab-scale PSA system with two beds
of ~250 g adsorbent each. The relevant inputs are summarized in the MainSpecs
table for the ACM block, which is presented in Table 2.
Table 2. MainSpecs table for the example provided.

Results of the PSA separation predicted by the above equations are presented in
Table 3. The model predicts a 25% recovery of the light key (O2) in the HPP, at an
elevated concentration of 61 mol%. Some of the results are different from Sees et
al. due to modification of the unity bed-state efficiencies assumed in the example.

7
Table 3. Separation results from the PSA example.

For an alternative solution, the recovery and purity can be set to Fixed, while PH and
mads are set to free, as shown in Table 4. This could be used to design a system for
a desired separation.
Table 4. Alternate solution with fixed purity and recovery.

4 Conclusion
This ACM model presents a method of representing a PSA column, a cyclic non-steady
state process, with a simplified steady-state approximation. When validated against

8
plant data and/or rigorous PSA simulations (for example, in Aspen Adsorption), the
holdup-based model described here can be used as a shortcut PSA. Additional future
enhancement could include:

 Heat duty calculations (based on isothermal operation)

 Adiabatic operation, incorporating heat of adsorption/desorption

 Estimates of compression power given source and venting pressures (i.e.,


ambient conditions)

9
References
[1] M. D. Sees, T. Kirkes, and C.-C. Chen, “A simple and practical process modeling
methodology for pressure swing adsorption,” Comput. Chem. Eng., vol. 147, p.
107235, Apr. 2021, doi: 10.1016/j.compchemeng.2021.107235.

10

You might also like