Intelligent modeling of the hydrogen sulfide removal by deep eutectic solvents for the environmental protection

Abstract

Hydrogen sulfide (H2S) is an acidic gas which produces by both natural (hot spring, volcano, microorganism decomposition) and industrial (biogas production, natural gas, liquefied petroleum gas, and oil refineries) processes. Due to the highly toxic and corrosive nature of H2S, researchers continuously search to find a practical scenario for its elimination from gas and liquid mixtures. Hydrogen sulfide absorption by deep eutectic solvents (DES), has recently found considerable attention, mainly in laboratory-scale investigations. Despite a relatively comprehensive experimental analysis of H2S absorption by DESs, there is no reliable empirical or intelligent methodology to predict the amount of this removal. Thus, the current study utilizes three machine learning classes (i.e., artificial neural network, support vector regression, and hybrid neuro-fuzzy system) to estimate the amount of H2S removal by 18 different DESs. The DES composition and operating conditions are the independent variables of these intelligent tools. The relevancy analysis approves that the amount of absorbed H2S by DESs is mainly controlled by hydrogen bond acceptor type, pressure, and temperature. Statistical and graphical analyses approve that the hybrid neuro-fuzzy system (hybrid NFS) is the highest accurate model to estimate the DES-H2S phase equilibria. The hybrid NFS predicts 495 experimental samples of H2S removal by DESs with the AARD = 6.11 %, MAE = 0.064, RAE = 5.64 %, MSE = 0.0149, and R-value = 0.99709. The combination of [C1-TMHDA]Ac and acetamide (1:2) is the best DES to achieve the maximum level of H2S elimination.