Research digital skills training 2021
Modelling dual reflux pressure swing adsorption (DR-PSA) units for gas separation in natural gas processing
George Zhang, Department of Chemical and Materials Engineering
Pressure swing adsorption is an alternative gas separation method using adsorption to separate gas mixtures. DR-PSA is an advanced configuration of pressure swing adsorption with two product refluxes and an internal loop. DR-PSA can achieve perfect separation in theory.
Dynamic modelling of adsorption units is considerably hard. Material balance, energy balance, heat transfer, mass transfer has to be considered in dynamics, thus rigorous numerical models are often built using a dedicated simulator: Aspen Adsorption. Aspen Adsorption is a module of the Aspen Suite and has built-in partial differential equations for mass balance, energy balance as well as adsorption isotherms. Besides these, Aspen Adsorption can model pressure-flow networks in a rigorous way.
Conventional PSA usually contain two columns working in parallel – one in adsorption mode and the other in regeneration mode. To simplify the model in Aspen Adsorption, it introduced an “interaction unit” to decrease computational intensity.
However DR-PSA has two refluxes and four adsorption parts thus four adsorption columns need to be independently and simultaneously simulated while the pressure-flow relationships of the system also has to be guaranteed. Besides these, the stiff manner of the system requires the integrator step size to be very small (as low as 1E-5 seconds), otherwise the integrator would either crash or result in a wrong predicted value.
Figure 1: The model built in the Aspen Adsorption for simulation
Figure 2: The experimental 2-column setup
This research is performed in collaboration with the University of Western Australia, where they perform the experiments. Figure 3 below shows the results of a temperature distribution comparison between experiment and simulation. There is a very good agreement, which could only be obtained thanks to the computational power provided by the VM.