Determination of Triethylene Glycol Concentration in Natural Gas Dehydration Systems Using Support Vector Machine Algorithm

Abstract

Triethylene glycol (TEG) liquid stream is used in natural gas dehydration units in order to prevent gas hydrate formation, the blockage of transportation pipelines, and liquid water condensation during transmission of natural gases in pipelines. Hence, it is needed to propose a reliable model to estimate the properties of TEG for better understanding of its performance. In this study, the robust least square modification of support vector machine (LSSVM) methodology is implemented to propose a computer program, by which the TEG concentration in natural gas dehydration unit can be determined at various reboiler pressure and temperature. In the final analysis, the Leverage approach (Williams plot) is applied to determine the applicability domain of the model and to detect any probable erroneous data points. It is found that the proposed TEG model is capable in low pressures compared to high ones. Additionally, the developed LSSVM model results indicate 0.08% average absolute relative deviation from the corresponding TEG concentration literature values, and squared correlation coefficient of 0.98.