(31d) Simulation of Near-Optimal and Valve-Free Operational-Half of Pressure Swing Adsorption Processes for Fair and Facile Evaluation of Process Performance Metrics

In this contribution, we will present a method for simulating near-optimal and valve-free operational-half of pressure swing adsorption (PSA) processes based on our new modeling and simulation framework, so called the valve-free model. The framework automatically determines near-optimal operational decisions during dynamic simulation of a PSA process. Also, the model eliminates any necessity for the user to specify inputs related to the detailed operation of a PSA process. Therefore, by using the framework, any non-expert users can easily obtain simulation results corresponding to a near-optimal operation of a PSA cycle.

First principle dynamic modeling and simulation of PSA processes has widely been used to capture dynamics of non-equilibrium adsorption phenomena. Furthermore, the approach has potentials to be utilized in many industrially relevant applications such as adsorbent screening, conceptual process design, optimal control, and optimization. Despite these potentials, however, there have been several limitations that prevented widespread adoptions of the first principle modeling and simulation in providing solutions to the aforementioned applications.

One of the major limitations of the current standard dynamic model, i.e. valve model, is its mathematically intractable formulation. For instance, in order to mathematically describe an adsorption column, the model needs to be formulated using a coupled system of nonlinear partial differential equations. Typically, solution methods to a such model are quite involved in terms of numerical methods used, conceptually hard to understand, and difficult to implement.

Another major limitation of the valve model is related to the user experiences in utilizing the model. Currently, to simulate a near-optimal PSA process using the valve model, the user has to supply a carefully chosen set of input parameters for a given PSA process simulation. However, choosing a set of input parameters that leads to near-optimal process simulation is not so simple as there practically exists so many reasonable ways to specify inputs such that the final simulation results are nowhere near-optimal.

Lastly, one of the most critical limitations of the valve model is the apparent sensitivity of simulation outputs on user supplied inputs. When designing chemical process systems, having accurate process performance metrics is considered as an essential requirement in determining process feasibility. However, due to the sensitivity issue, performance metrics obtained by the valve model are substantially influenced by the user supplied inputs. Furthermore, identifying an optimal set of inputs for valve model is computationally exhausting as it requires time consuming trial-and-error simulation studies.

Therefore, these limitations prevented a broader group of researchers and practitioners from being able to accurately predict performance metrics related to PSA process. Since enabling chemical industry to reliably assess adsorption based process intensification (PI) concepts would lead to a faster and more widespread deployment of such novel technologies, there is a critical need for developing a new dynamic model for simulating near-optimal PSA processes.

To address this critical need, a novel valve-free model has been developed to enable the user to simulate near-optimal PSA processes with Skarstrom type PSA cycles. In order to develop the valve-free model, the Skarstrom PSA cycle was decoupled into two halves including different steps, namely the operational-half and re-generation-half. Based on the simplified system, near-optimal operational policies for the two halves were devised separately.

A near-optimal operational policy for the re-generation-half was presented in previous talks given during the 2019 Spring AIChE meeting^[1] as well as the 2019 Annual AIChE meeting^[2]. Therefore, in this oral presentation, we are going to discuss the technical details that are involved in the development of a near-optimal operational policy for the operational-half.

In order to develop the valve-free model, several key assumptions were made to simplify the system and to present a proof-of-concept for the proposed modeling framework. The current formulation of the valve-free model assumes isothermal, no dispersion effects, binary ideal gas mixture, linear isotherm, and linear driving force adsorption kinetics.

The power of the proposed methodology (i.e. valve-free model) lies within enabling the user to run high quality PSA process simulations without having to preconceive the respective optimal operational policies. This enhanced user experience may lead to more wide adoption of process simulation of PSA processes as a tool that can help make critical day-to-day decisions or assessments that are related to operation, control, and optimization of PSA processes.

For the scope of the future research goals, the merge between the decoupled halves will be sought after in hoping to devise an overall near-optimal operational policy for PSA processes involving Skarstrom type cycles. Sequentially, based on the near-optimal operational policy for PSA processes generated by our valve-free model, further interrogations on the development of design rules for synthesizing near-optimal PSA processes will be made.

Ultimately, by serving as practical tools that can address the identified knowledge gap, valve-free model can potentially contribute to the advancement of PI and deployment of various adsorption technology in the chemical process industry.

^[1] Kim, T.-H. et al. (2019). Rational Screening of Adsorbents for Natural Gas Upgrading By Pressure Swing Adsorption Using Dynamic Simulation of Process Performance Metrics, April 2, 2019, Spring AIChE Meeting, New Orleans, LA, 2019.

^[2] Kim, T.-H. et al. (2019). A 'Valve-Free' Model for Dynamic Simulation of Pressure Swing Adsorption Processes That Automatically Determines Near-Optimal Operational Policies, November 11, 2019, Annual AIChE Meeting, Orlando, FL, 2019.