Download Full Text (739 KB)


Liquid loading is a term used to describe the situation where the gas produced from a well is unable to carry the liquid that is also produced along with it (Khetib et al., 2022). As a result, the liquid starts to accumulate in the wellbore. This accumulation of liquid can cause a decrease in gas production and in severe cases, it may even lead to a complete stoppage of production (Khetib et al., 2022, 2023) . The phenomenon of liquid loading in gas wells occurs when the critical gas velocity is less than a certain value, leading to a decrease in gas flow rate and ultimately a decrease in production (Merzoug et al., 2022). To simulate this phenomenon and study it in detail, we conducted experiments in a multiphase flow loop. We aimed to compare the results of the experiment with machine learning algorithms to predict the loading and unloading of the well. The purpose of this experiment was to examine the start of liquid accumulation in a gas well using a 2.4 meter vertical rigid pipe system with a 0.0508 meter (2 inch) internal diameter (Khetib. Y, 2022). The study analyzed the flow of gas and liquid in a vertical direction to gain insight into how liquid builds up in a vertical tube as gas flow decreases. We varied the gas and liquid flow rates to simulate different conditions and recorded the pressure and flow rate data (Khetib. Y, 2022). We used this data to train machine learning algorithms such as Support Vector Machines (SVMs) (Ifrene et al., 2023), Random Forests, XGBoost, and Neural Networks, to predict whether the well is loaded or unloaded. We then compared the predictions of the machine learning algorithms with the experimental data.

Publication Date



Bismarck, ND


Adapted from extended abstract based on oral presentation given at 2023 AAPG Rocky Mountain Section Meeting, Bismarck, North Dakota, June 4-6, 2023.

Machine Learning Algorithms for Predicting Liquid Loading in Gas Wells