Soft Sensing in Deep Wells Within Artificial Gas Lift Technology

Abstract

Artificial gas lift technology is used in the oil industry to enhance oil recovery. Controlling this process is considered as a vital task by many researchers in the field. Solution of this problem would improve process performance and stabilize system operation. However, most of the control methodologies found in the literature and oil-field practice use only such process variables as on-surface flow and pressure measurements. The proposed approach offers soft sensing of process variables in the well that are unavailable for measurement due to technical constraints. For instance, measuring the flow rate of gas or liquid at the depth of a few kilometres would be unfeasible, given that the existing instruments cannot endure huge pressures and high temperatures at this depth. The proposed methodology provides soft sensing of these variables by utilizing a sliding mode observer (state estimator). Sliding mode observers are widely used for acquiring good estimates of unmeasured process variables in a dynamical system. This proposed soft sensing is primarily intended for control purpose. For that reason the estimator must provide computational efficiency with acceptable rate of accuracy, since the observer must be working in real-time to be a part of the control system. By linearizing the nonlinear model of the artificial gas lift dynamics, we reduce the complexity of implementation to simple arithmetic operations that must be realized within an estimator. In this case, computational efficiency of these operations would be higher than that involving solution of complex functions and algebraic equations that are involved with the nonlinear model of the gas lift. Therefore, the proposed linear estimator is simple, higly efficient in term of computation, and it satisfies the requirement for real-time operation. We developed a simple version of sliding mode observer through linearization principle applied in a number of points. The highly nonlinear gas lift system was linearized at different points which were selected as corresponding to different opening of the valve. This method is applicable for this scenario as the linearization of the model is valid around each point. Incorporating multiple sliding mode observers was achieved by utilizing the interpolation principle. This novel approach was tested with the nonlinear gas lift model and simulation results was obtained for 10-point and 20-point observers. The resulting estimates tracks the actual values with insignificant error. By checking the percentage error of our estimates, we conclude that the designed observer provides good and reliable results. This would permit further enhancements in the gas lift technology with less cost, by utilizing these estimates to control the nonlinear process using some advanced control techniques such as model predictive control.

Multiple‐model sliding mode observer for an artificial gas lift system

“…Therefore, first‐order conventional sliding mode observers are used to produce estimates of the deviations, where instead of eigenvalues we consider an observer gain matrix. The detailed design strategy for a single‐point sliding mode observer was presented in for a linearized system at a specific operating point. The current paper describes the design strategy of a multiple‐point sliding mode observer for the nonlinear artificial gas lift system.…”