Prediction and Optimization of Gas Lift Performance Using Artificial Neural Network Analysis

Abstract

Gas lift is one of the most widespread methods of artificial lift technologies used when wells’ production rate drops below the economic limit. Gas Lift is employed to maintain the production above the available limit by means of injecting gas into the tubing through the casing–tubing annulus and a gas lift orifice installed in the tubing. Gas lift has been widely used in the oil fields that suffer from sand production. It is also used in deep and deviated wells and on offshore platforms. Lifting costs for a large number of wells are generally low. However, capital costs of compression stations are very high, so it is necessary to optimize gas lift wells by determining the optimum gas lift injection rate and optimum oil rate for each well. In this paper, conventional nodal analysis models using Pipesim software were used to predict the optimization parameters based on wells flowing survey, reservoir and well parameters and calculations of multiphase flow behavior. Artificial neural network (ANN) models were also used based on gas lift databases and gas lift monitoring systems. ANN models were trained to obtain the optimum structure and then tested against pipesim models. Also, this paper presents a new theory about the relative importance of gas lift system input data in predicting optimum parameters of gas lift system. It has been concluded that ANN has an excellent competing ability for gas lift optimization prediction compared to conventional methods and can be used interchangeably. This technique can considerably help in the immediate optimal design of gas lift wells.