Model predictive control of artificial gas-lift process

Abstract

Artificial gas-lift (AGL) technique is one of the most widely used methods in the oil production to maintain acceptable oil flow to the processing equipment when the reservoir pressure is not high enough. In spite of its popularity, the AGL process is prone to casing-heading instability phenomenon, which is revealed as significant flow oscillations. This is undesirable as it results in production losses and unstable behavior that has negative impact on the downstream equipment. Controller design for such a process is very challenging as it exhibits highly nonlinear dynamics. In this thesis, we investigate the application of different types of model predictive control (MPC) techniques to the AGL. First, a new design of subspace predictive controller (SPC) is developed for gas-lift process. The SPC is a data driven algorithm, which is based on linear predictor to predict future output based on process input and output data. The linear prediction model is derived offline. Thereby, the key feature of the proposed approach is that precise knowledge of the model and on-line optimization are not required to derive the control law. Then, another type of MPC, the predictive generalized minimum variance control (PGMVC) is employed to derive a robust controller based on the state estimation to stabilize the AGL process when casing-heading phenomenon occurs. A closed-form optimal control law is obtained based on the Taylor series approximation. Further, a nonlinear state observer is produced and combined with the controller to ensure closed-loop control through variables that are most beneficial to the system performance, which are unmeasurable and can be obtained only via estimation. Through simulation studies, the effectiveness of the proposed controller is demonstrated. Finally, the stability analysis of AGL process is discussed by operating both openings of gas injection choke and oil production choke. Furthermore, a trajectory optimization algorithm is developed to minimize the process transition time, where the operation point moves from unstable region to stable region.

Date of Award	2016
Original language	American English
Supervisor	Ahmed Al Durra (Supervisor)