Building pressure swing adsorption model for H2 purification

Driven by the Paris Agreement signed in 2015, Shell PLC have developed its own schedule and objectives to become Net-Zero emission by 2050.
Hydrogen as energy carrier, plays a crucial role in the energy transition process. Depending on the source, hydrogen is classified as high carbon, low carbon or renewable, among others. In addition, the production process will determine the contaminants present in these H2-rich streams. Hydrogen is used in a wide range of applications which demand different levels of purity, from above 98% to 99.99999% mol/mol.
Pressure Swing Adsorption (PSA) is proven to be the most cost-effective technology to reach these purity qualities at industrial scale. PSA capacities go from 1000 Nm3/h to 400.000 Nm3/h. Therefore, it is of great importance for Shell’s business to have in-depth knowledge about this unit. For this reason, Shell wish to develop a PSA model for hydrogen purification applications.
The first step to build this model consisted in defining the stakeholder’s requirements related to the features that should be included in the model. From this stage, it was concluded that a tool needed to be develop in a Microsoft Excel-MATLAB platform, since it combines a known software like Excel with the calculation capability of MATLAB. The decision was made based on finding an “user-friendly” software with the power of solving the complex partial differential equations that govern the physics present in the adsorption phenomena.
Secondly, the “archetypes” of feed streams and product streams for this unit were defined, along with the most commonly used adsorbent materials based on data found in literature and operating plants within Shell. This step was used to define the components and adsorbents to be included in the tool and establish its range of applicability.
Then, two separate tools were built. The Preliminary sizing tool (Excel) included all the necessary equations and design criteria to size a PSA unit and define the adsorption cycle. Later, the Mass balance estimation tool (MATLAB) was created to perform mass balance calculations using the results given by the Preliminary sizing tool. In the final stage, both tools were combined into a single PSA modelling tool which offers two operational modes: Design mode and Simulation mode.
As final development stage, the PSA modelling tool results were compared against operating data to tune its performance. The tunning procedure was done taking into account two different PSA applications.
To determine the Technical Design accuracy of the tool, the results given by it were compared with additional operating data and the average relative error for both modes was calculated. The Design mode had an average relative error of 10%, meanwhile, the Simulation mode had an average relative error of 14%, giving an overall average technical error of 12%. The Cost Estimation predictions given by the tool have an error of +/- 50% (order of magnitude estimate). Both accuracy levels were considered acceptable and sufficient by Shell’ stakeholders. The PSA modelling tool included a manual and video tutorials as part of the final tool.