Title: Predictive Modeling of Glycol Recovery Efficiency in Natural Gas Processing Using Machine Learning
Authors: Christian Emelu Okalla, Charley Iyke C. Anyadiegwu, Ubanozie Julian Obibuike, Prospere Chimdike Nwokorie, Chukwudi Michael Ohaegbulam, & Chika Veronica Nwachukwu
Volume: 10
Issue: 2
Pages: 78-84
Publication Date: 2026/02/28
Abstract:
Gas dehydration is an important aspect of gas processing since it tends to remove water from natural gas there-by minimizing the formation of gas hydrates during transportation in pipelines. Gas dehydration is usually conducted using triethylene glycol (TEG) which is usually recovered after the process. It is important to note that the recovered TEG should have a high purity so that when reused, it can be more effective. A solution to this problem is proposed in this study such that a model is developed which can be used to predict TEG purity and hence define specific input values to the system to achieve the desired purity. The model was developed using Design Expert Software with input data from design of experiments and output data from Aspen Hysys simulation of TEG purity. Analysis of Variance was conducted to determine the model significance and model terms that are significant to the model. The developed model was validated using cross plots and trend analysis which proved to be valid. This is because the data points were located on the 45° line and also the actual and predicted results followed the same pattern depicted by overlap of the curves. The model developed in this study can be used to predict TEG purity for given values of TEG flow rate, column pressure, and reboiler temperature. This work can aid in reducing costs associated with performing trial and error simulations with the real system or with the simulation software.