Implementation of Artificial Neural Networks in the assessment of CO2 solubility in deep eutectic and ionic liquid solvents – Performance and cost comparison

Abstract

Background

In order to counteract the economic and environmental issues presented by Ionic Liquids (ILs) for carbon capture in post-combustion processes, Deep eutectic solvents (DESs) are being researched as potential absorbents. These are an emerging class of ILs that have a strong contribution from hydrogen bonding and have shown promising trends in CO₂ absorption in recent times.

Methods

In this study, three hydrogen bond acceptors (HBA), along with 2-hydroxypropanoic acid (Lactic Acid (LA)) as the hydrogen bond donor (HBD), have been identified and analyzed as CO₂ absorbents. Considering their structural properties, thermodynamic behavior, and experimental conditions as input parameters, a backpropagation neural network (BPNN) has been implemented to analyze and predict the extent of CO₂ solubility within each of the DES mixtures.

Significant Findings

BPNN successfully predicted trends in the solubility as a function of the alkyl chain length, temperature, and pressure. It was observed that the solubility of CO₂ increased with increasing alkyl chain length and pressure but decreased with increasing values of temperature. DES is found to be more economical than other ionic liquid solvents used for CO₂ absorption.