Simulation and optimization of an industrial PSA unit

Abstract

The Pressure Swing Adsorption (PSA) units have been used as a low cost alternative to the usual gas separation processes. Its largest commercial application is for hydrogen purification systems. Several studies have been made about the simulation of pressure swing adsorption units, but there are only few reports on the optimization of such processes. The objective of this study is to simulate and optimize an industrial PSA unit for hydrogen purification. This unit consists of six beds, each of them have three layers of different kinds of adsorbents. The main impurities are methane, carbon monoxide and sulfidric gas. The product stream has 99.99% purity in hydrogen, and the recovery is around 90%. A mathematical model for a commercial PSA unit is developed. The cycle time and the pressure swing steps are optimized. All the features concerning with complex commercial processes are considered

CO2-Argon-Steam Oxy-Fuel Production for (CARSOXY) Gas Turbines

“…However, the typical process which is followed in the industry to separate H 2 from the WGS products is PSA. References [16][17][18] indicate that PSA is insensitive to the changes in molar fractions of the feed stream or to its boundary conditions (i.e., temperature or pressure), where 90% of the hydrogen is recovered with a 99.99% purity. A detailed description of the process is available somewhere else [17].…”

Design and simulation of a petcoke gasification polygeneration plant integrated with a bitumen extraction and upgrading facility and net energy analysis

“…The gasification technology selected for this study is the Shell/Prenflo gasifier, for which Suncor's petcoke is considered to be an ideal feedstock [43]. The technology consists of a dry feed upflow reactor with a membrane wall vessel.…”

“…In the CO 2 venting case, to complete the desulphurization process, the COS should be converted into H 2 S before the acid gas removal stage. The COS can be selectively converted into H 2 S through a catalytic hydrolysis reaction [42,43]. The hydrolysis reactor has been modeled as an RGibbs reactor with restricted equilibrium where only the COS hydrolysis reaction is allowed.…”

“…Given the complexity of the involved physical/chemical phenomena and the availability of literature data, the air separation unit (ASU), the acid gas removal (AGR) process and the CO 2 compressor were not modelled in detail but their performance (energy consumption and separation efficiency) were taken from the work by Martelli et al [43] who considered a similar Shell-based IGCC with/without CO 2 capture and storage, with a net electric production capacity within the range of 270 MW -315 MW. Without loss of accuracy, the power consumption of the ASU (providing 95% purity O 2 ) is assumed to be proportional to the mass flow rate of produced oxygen (2.482 MJ/kg O 2 ).…”

Analysis of the boundary conditions for the simulation of the pressure equalization step in PSA cycles

“…Although the pressure equalization step is usually included in multibed PSA cycles and plays a key role in the minimization of the energy consumption of the cycle (this step is included to conserve energy and separative work), the simulation of this step has not been widely investigated up to date, specially the aspects concerning the boundary conditions. After reviewing the literature about dynamical simulation of the pressure equalization step, one finds that two different kinds of boundary conditions (which are not equivalent) are proposed to simulate the time-profile of the dependent variables in the bed ends that are connected to each other: (a) the well-known Danckwerts boundary conditions (DBC) without modification (Malek and Farooq, 1997;Yang et al, 1997;Jee et al, 2005;Kim et al, 2006) and (b) modified DBC, replacing the original equations for the fluid entering a bed by an equalization of mole fractions and temperature, i.e., the mole fractions and temperature of the bed receiving flow are made equal to ones corresponding to the outlet of the source bed (Barg et al, 2000), or only the mole fractions for an isothermal model (Mendes et al, 2001). All these studies use the dynamic model to simulate experimental data.…”