(462e) Machine Learning Identified Process-Performance Limits of Solid-Sorbent Direct Air Capture (DAC)

In order to limit the global surface-temperature increase to 1.5°C, the recent IPCC report concludes that net–zero CO₂ emissions have to be achieved by 2050, thus highlighting the importance of negative emission technologies such as direct air capture (DAC) [1]. This work deals with adsorption-based DAC, particularly a steam-assisted temperature vacuum swing adsorption cycle (s-TVSA) process. We ask the key question, “What are the performance limits of such a process, provided the ideal adsorbent can be synthesized?” To perform the process performance limit study, the CO₂/N₂ adsorption isotherms, material properties such as particle density and specific heat capacity of the sorbent alongside the cycle operating conditions are treated as variables in our approach. Holding the isotherms to be decision variables results in an adsorbent-agnostic approach. Using the detailed process model for the performance limit studies is computationally expensive; hence, machine learning models have been developed in this work. The advantages of such a machine-assisted adsorption process learning and emulation (MAPLE) framework for the PVSA cycle in post-combustion CO₂ capture have been shown by Pai et al [2,3]. Sample datasets are generated using the detailed model. The machine learning toolbox available in MATLAB is used and Artificial Neural Network (ANN) models have been developed to accurately predict the Pu, Re, En, and Pr. ANN models are validated against the detailed models by comparing the results for a select few sorbents. Numerical optimizations are performed with the ANN models developed for the TVSA cycle to study what properties would constitute the “ideal sorbent”.

The MAPLE model is complemented by a technology-agnostic simplified analysis of DAC. The simplified analysis provides the bounds for En requirement under both dry and humid conditions. The simplified analysis provides insights into the key principles that control the process performance and can be used to identify potential adsorbents for process-level studies.

References:

1. IPCC, Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change., 2021
2. K. N. Pai et al., “Practically achievable process performance limits for pressure – vacuum swing adsorption-based post combustion CO2 capture”., ACS Sustainable Chem. Eng., 2021, 9, 10, 3838-3849.
3. K.N. Pai et al., “Generalized, Adsorbent-Agnostic, Artificial Neural Network Framework for Rapid Simulation, Optimization, and Adsorbent Screening of Adsorption Processes”., Ind. Eng. Chem. Res., 2020, 59, 38, 16730 - 16740