Prediction of gas-lift performance using neural network analysis

Abstract

<abstract> Integrated gas lift system optimization plays an indispensable role in production and economics by maximizing the revenue from a gas lift field. This requires: optimization of gas lift parameters, finding the best tuning of the completion and surface production systems to keep pace with the dynamic reservoir changes along with saving gas quantities and compression costs. Accordingly, a comprehensive study is being carried out to measure the capability of the Artificial Neural Network (ANN) and Machine Learning (ML) in the optimization of gas lift parameters. The results of this study show the power of two different mechanisms of neural network (NN) which are Radial Base Function (RBF) and Back Propagation Function (BPF) to predict the most three important factors of the process: optimal gas injection rate, bottom hole pressure and flow rate and compare the findings with conventional methods. In addition, this work provides 3 functional equations that can be utilized by applying the field data with no artificial intelligence (AI) expertise or software knowledge. This effort provides forth an industrial insight into the role of data-driven computational models for the production recognition scheme, not only to validate the well tests, but also to reduce the uncertainties in production optimization. The work was completed by generating an economic analysis to illustrate the understanding of potential benefits of implementing irregular gas lift mechanisms in the field to stand on both technical and economic aspects of the study. </abstract>