(348a) A Procedure for Identifying a Nearly Pareto-Optimal Operation Policy for Pressure Swing Adsorption Processes

In this research poster presentation, we describe an efficient and systematic procedure for identifying nearly Pareto-optimal operational policies for pressure swing adsorption (PSA) processes. Under the new procedure, a complicated decision making process for PSA process operation can be simplified and the user can conveniently make operational decision accounting for a tradeoff relationship between the cycle duration and a total amount of product produced in a PSA cycle. This improved decision-making procedure substantially reduces the number of fine tunings required by a trial-and-error procedure which is a key debottleneck in utilizing simulations for designing PSA processes.

Initialization of an instance of PSA process simulation is a non-trivial task to perform due to a plethora of simulation input parameters that need to be specified. Fortunately, measurable inputs like adsorbent physical properties and characteristics can be obtained from independent experiments. In contrast, inputs for determining operation policy and a system design must be decided carefully just to have a feasible PSA process. Operation policy is determined by inputs such as the duration and boundary volumetric flow rates for each step in a PSA cycle. Also, design decisions for PSA processes is typically specified by the aspect ratio, the pressure ratio, and the feed to purge ratio.

While the inputs for making operational decisions are intuitive in their meanings, specifying their actual numerical values is not necessarily intuitive. Furthermore, PSA process simulation outputs can be very sensitive to the inputs, demanding operational and design decisions to be made optimally. Naturally, mathematical programming technique can be applied to make decisions optimally by solving an optimization problem. However, process optimization does require expertise from the practitioner and decision makers without such backgrounds will have hard time making decisions solely based on the optimization results. On the other hand, the user can continuously refine the decisions for operation and design by implementing a trial-and-error fine tuning process. Again, a such procedure may require a significant number of refinement operation and can be prohibitable in making optimal decisions quickly enough.

To address this critical need, we developed valve-free operation policy capable of elucidating a tradeoff relationship between product throughput and productivity for a given PSA process. The valve-free policy can generate a Pareto optimality plot for a given PSA cycle by considering the following insights. First, by controlling each relevant boundary volumetric flow rate as a function of state variables, the driving force for desorption during re-generation half is maintained in an optimal way. Second, based on equilibrium theory pre-calculations, a set of useful normalization constants are derived and utilized to reparameterize the dynamic model in terms of more intuitive dimensionless parameters.

To identify a Pareto optimal operation policy, the user can choose a Pareto optimal point from the Pareto chart and translate a set of corresponding valve-free parameters into a set of actual input parameters for a given PSA process. The user can specify duration of each step and associated valve constants based on valve-free operation policy so that a PSA cycle reflecting a desired tradeoff relationship can be simulated with a high fidelity. Once this is done, the user can run a process simulation until a cyclic steady state (CSS) convergence is attained. At this point, the process performance metrics can be calculated reliably with a substantially reduced number of trial and error procedures.

In conclusion, we believe that the systematic procedure discussed in this poster presentation will be an indispensable tool for any PSA process simulators in terms of helping the user to make operation and design decision optimally. Thanks to this advancement, we expect to see more practical PSA process designs that are originating from simulation tools in conjunction with experiments rather than solely based on experimental efforts.