CONSTRAINED OPTIMIZATION IN GAS LIFT TECHNIQUE FOR OIL PRODUCTION WELL

• Main Author: SAEPUDIN (NIM: 30104005), DENI
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/11980
• Institution: Institut Teknologi Bandung

A mathematical model arising from oil production process using Gas Lift technique is discussed in this thesis. This technique is an artificial lift method commonly used to increase oil production rate from oil producing wells those produce oil in low level rate, or from oil production wells those cannot produce oil by natural flow method.<p>Here, Gas Lift problem is formulated in a mathematical model as a boundary value problem. The model is obtained by combining reservoir fluid flow and tubing fluid flow. The model relates some parameters which are going to study, especially parameters gas injection rate and liquid production rate. Relationship between these two parameters could be described using a Gas Lift Performance Curve.<p>A new form of constrained optimization problem is obtained through this study. The liquid production rate will be maximized subject to the boundary value problem constraint of the Gas Lift model. This is a new point of view and lead to a new approach in solving Gas Lift optimization problem in simpler, more flexible and integrated way.<p>This approach is applied to solve the optimization problems, starting from a single Gas Lift well problem and then for a Dual Gas Lift well with considering stability aspect, and finally for a multi Gas Lift wells system with gas availability constraints and surface facilities constraints. The constrained optimization problem could be transformed to an unconstrained optimization problem using penalty approach. A computation scheme has been developed to solve the optimization problem. Considering complexity of the problem, the computation scheme is developed based on genetic algorithms. Computation results show quite good estimation for optimum solution. The approach also gives better quality prediction over existing approach, since all computation results come from the model, not from regression or interpolation. <br />