Machine learning-based models for estimating the rich amine acid gas loading based on industrial data

To efficiently design the processes dealing with acid gas absorption, it is crucial to have robust and accurate predictive tools for acid gas loading within amine solutions. In this study, industrial data collected from Ilam Gas Treating Company, Iran, were utilized to model the acid gas loading in rich alkanolamine solutions. The collected data cover diverse ranges of operating factors, such as the flow rate and temperature of acid gas flow, as well as the concentration and temperature of the amine solutions. To derive reliable models, artificial intelligence approaches, such as multilayer perceptron neural network (MLP-NN), Gaussian process regression (GPR), and radial basis function neural network (RBF) were employed. According to statistical analysis, all intelligent models generated reliable predictions for acid gas loading, since the average absolute relative errors (AAREs) of 2.71%, 3.58%, and 5.11% were obtained by MLP-NN, GPR, and RBF models, respectively. Moreover, the newly proposed models benefited from a high accuracy level, and the majority of their estimations fell within ±5% errors from the actual values. The individual and combined effects of operational factors on the acid gas loading in rich amine solutions were investigated in detail, and the most important factors were determined based on a sensitivity analysis.