Comparison of SQP and AL algorithms for deterministic constrained production optimization of hydrocarbon reservoirs

Reservoir production optimization involves the determination of well control settings which maximize a pre-defined reservoir production preference such as ultimate oil recovery or net present value (NPV). Honoring the nonlinear constraints, such as producer's water cut and bottom hole pressure...

Full description

Saved in:
Bibliographic Details
Published in:Journal of petroleum science & engineering Vol. 171; pp. 542 - 557
Main Authors: Liu, Zhe, Forouzanfar, Fahim, Zhao, Yu
Format: Journal Article
Language:English
Published: Elsevier B.V 01.12.2018
Subjects:
Adjoint gradient
Augmented Lagrangian
Net present value (NPV)
Nonlinear constrains
Production optimization
SQP
Net present value (NPV)
Production optimization
Nonlinear constrains
SQP
Adjoint gradient
Augmented Lagrangian
ISSN:0920-4105, 1873-4715
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Reservoir production optimization involves the determination of well control settings which maximize a pre-defined reservoir production preference such as ultimate oil recovery or net present value (NPV). Honoring the nonlinear constraints, such as producer's water cut and bottom hole pressure (BHP) constraints for rate specified wells, is the most challenging part of the production optimization. This work entails a comparative study of the augmented Lagrangian (AL) and sequential quadratic programming (SQP) algorithms for constrained reservoir optimization. Both methods have been widely used to solve the nonlinear constrained optimization problems in the numerical optimization community. The optimization of PUNQ reservoir model is considered in order to compare the performance and robustness of the AL and the SQP, where the gradients of the objective function and nonlinear constraints are estimated using the adjoint method. The computational results showed that with properly tuned parameters and good initial points, the AL could yield a slightly higher ultimate NPV in some cases. However, the SQP generally has better performance regarding efficiency, robustness and constraints handling. For example, under the SQP generated optimal well controls, the watercut of each producer would strictly obey the constraint, while the optimal well controls generated by AL would lead some minor valuations. Moreover, an SQP-AL algorithm is also proposed to solve the constrained production optimization problem by applying the AL search direction on the convergence of SQP. The results showed that the SQP-AL algorithm is able to achieve a higher NPV with slightly more iterations compared with the SQP method. •A comparative study of the AL and SQP algorithms for constrained reservoir optimization.•Compare the performance of SQP and AL regarding efficiency, robustness and constraints handling.•Propose the SQP-AL algorithm by applying the AL search direction on the convergence of SQP.•Implement the adjoint gradient for 3D 3Phase general reservoir simulator.
ISSN:0920-4105
1873-4715
DOI:10.1016/j.petrol.2018.06.063