Reducing Gas Well Uncertainties by Predicting Liquid Loading Using Artificial Neural Network

Reducing Gas Well Uncertainties by Predicting Liquid Loading Using Artificial Neural Network

Liquid loading is an important issue caused by fluid accumulation in the tubing of gas wells when the gas kinetic energy is not sufficient to carry liquid slugs to the surface. This problem has influences onproduction capacity of gas wells; For example, in high-pressure wells, it disturbs well production byslugging and churning or in low-pressure wells, it may kill the well. Moreover, in reservoir engineering, the liquid loading may cause uncertainties in well test data. Despite the fact that there aresolutions for liquid loading such as gas lift or pumping, preventing it eliminates load up costs. Thebest way is to continue gas production at a flow rate above a critical value to prevent liquid loading. In this paper, we present a new method to estimate the critical flow rate as accurate as possible to predict the occurrence of loading in a gas well. In our approach, we use artificial neural network as afast, easy to learn, and reliable method to provide the results for production engineers. The developed network is trained and tested with available data from different gas wells. Our results are in good agreement with the field data and show less than 2.5% error in liquid loading prediction