Hybrid modeling of vacuum swing adsorption carbon capture process for rapid process-level evaluation of adsorbents

Adsorption of CO2 from stationary carbon sources represents a promising technology for carbon capture due to its intrinsically low energy requirement. While numerous highly selective adsorbents with good capacities are being developed and investigated, the evaluation of their true potential requires process simulation and optimization. These tasks are impeded by certain features intrinsic nature of the process, which involves solving a system of dynamic non-linear partial differential equations (PDEs) to reach cyclic steady state results in a timeconsuming task. Simpler models that can replace the complex model are needed for the quick evaluation of thousands of adsorbents in terms of their intrinsic energy requirements through the optimization of their process design and operation. First-principles should underpin the model to ensure accuracy and versatility so that the model can measure the true performance of the absorption system which uses various type of adsorbents in various conditions. In this study, we propose a hybrid modeling method for the vacuum swing adsorption (VSA) process, capable of predicting the energy consumption of the process and the bed profile using only tens of equations. This enables rapid process-level evaluations of candidate adsorbents. The method incorporates datadriven surrogate models into the first-principle model for the prediction of empirical parameters. The dataset for constructing the data-driven surrogate model is generated by simulating the rigorous first-principle process model, and the hybrid model is developed and validated using the dataset. To illustrate the use of this hybrid model, a case study involving two adsorbents is performed, including process evaluation, optimization, and sensitivity analysis. The case study also suggests key insights into the development of a VSA process.