A systematic review on the application of machine learning in carbon dioxide absorption in amine-related solvents

Abstract

Amine absorption has been regarded as an efficient solution in reducing the atmospheric carbon dioxide (CO2) concentration. Machine learning (ML) models are applied in the CO2 capture field to predict the CO2 solubility in amine solvents. Although there are other similar reviews, this systematic review presents a more comprehensive review on the ML models and their training algorithms applied to predict CO2 solubility in amine-related solvents in the past 10 years. A total of 55 articles are collected from Scopus, ScienceDirect and Web of Science following Preferred Reporting Items for Systematic Review and Meta-Analyses guidelines. Neural network is the most frequently applied model while committee machine intelligence system is the most accurate model. However, relatively the same optimisation algorithm was applied for each type of ML models. Genetic algorithm has been applied in most of the discussed ML models, yet limited studies were found. The advantages and limitations of each ML models are discussed. The findings of this review could provide a database of the data points for future research, as well as provide information to future researchers for studying ML application in amine absorption, including but not limited to implementation of different optimisation algorithms, structure optimisation and larger scale applications.