Title: Data-Driven Machine Learning Approach to Gas Lift Field Optimization in a Network of Clustered Wells
Authors: Ubanozie Julian Obibuike, Christian Emelu Okalla, Chijioke Ugochukwu Iyke-Onyeka, Favour Mesoma Onyema and Prospere Chimdike Nwokorie.
Volume: 10
Issue: 1
Pages: 10-18
Publication Date: 2026/01/28
Abstract:
Oil fields are usually produced using the energy present in the reservoir provided by drive mechanisms. When the energy of the reservoir is exhausted, artificial lift methods are used which can be gas lift or pumps. In this study, we look at gas lift. Most research have considered gas lift on a well-by-well basis which fails to address the dynamics of flow and also how each component of the production system affects each other. This study presents an approach for solving this problem by developing a model for predicting oil rate and bottomhole pressure which can be used in a field and the optimization conducted simultaneously. In this study, an oil rate and bottomhole pressure models were developed and validated using cross plots and trend analysis. Also, from cross plots, R2 values of 0.9953 and 0.9572 were observed. With the aid of genetic algorithm and defining a given number of wells in a field, the optimal settings for each gas lifted well can be determined using a multi-objective optimization algorithm such as multi-objective genetic algorithm. The results and approach used in this study can assist in solving challenges associated with developing a gas lifted field.