Neural networks versus Linear and Sequential Programming for Gas Lift Optimization in a two oil wells system

Using a model-based optimization, a neural network model is developed to calculate the optimal values of gas injection rate and oil rate of a gas lift production system. Two cases are analyzed: a) A single well production system and b) A production system composed by two gas lifted wells. The results were compared with the linear and sequential programming for gas lift optimization. For both cases minimizing the objective function the proposed strategy shows the ability of the neural networks to approximate the behavior of an oil production system and to solve optimization problems when a mathematical model is not available.