First-principles based surrogate modeling of pressure swing adsorption processes for CO2 capture

Pressure swing adsorption (PSA) is a highly versatile separate technology and represents a promising option for carbon capture. A dynamic model for a PSA process comprises partial differential equations, which are difficult to use for optimization tasks due to their long computational time and numerical instability problems. A surrogate model for PSA processes which captures the essence of the dynamics to predict key performance measures can enable process optimization to be carried out without the need to evaluate the rigorous model repeatedly. To this end, a surrogate model that predicts energy consumption, removal, purity, and CO2 avoidance cost for given operating conditions is developed as a system of nonlinear equations. The surrogate model is validated and can be used to find optimal operating conditions without having to carry out time-consuming dynamic simulations. It is expected that the ability of fast-optimization brings further opportunities such as screening of adsorbent candidates in material discovery researches.