Theoretical Research On Reservoir Closed-loop Production Optimization

Closed-loop production optimization combines the process of history matching and production optimization together to periodically updates the reservoir model and determine the optimal control strategy for production development to realize the goal of decreasing the knowledge of model uncertainty as well as maximizing the economic benefits for the expected reservoir life. The adjoint-based-gradient methods seem to be the most efficient algorithms for closed-loop management. Due to complicated calculation and limited availability of adjoint-based-gradient in commercial reservoir simulators, the application of this method is still prohibited for real fields. Thus, the paper mainly focused on the derivative-free optimization methods for closed-loop production management. In production optimization stage, a new derivative-free algorithm noted as QIM-AG was proposed for solving the optimal contol model of the reservoir production system. The algorithm was similar to the quasi-Newton method, which the objective function was optimized iteratively by sloving its quadratic approximation model. Based on the QIM-AG algorithm, the constraint production optimization and the robust optimization problem were also investigated. In history matching stage, the mathematic history matching model was built according to the Bayesian statistics theory. By using a set of unconditional realizations, a new parameterization method was proposed to decrease the dimension of the objective function to be minimized in history matching. Then the derivative-free algorithms combined with this parameterization method and the data assimilation methods were applied for history matching separately, which provided new approaches for large-scale automatic history matching problem. The results shows that the QIM-AG algorithm can be easily coupled with any commercial simulator without the calculation of the adjoint-gradient and outperforms the other derivative-free optimization algorithms considered with more computation efficiency and faster convergence speed. The QIM-AG algorithm combined with the Augmented lagrangian method is a good way to solve any constraints production optimization problem. The optimal estimated controls by robust optimization with QIM-AG algorithm results in a better robustness developing plan for geoloigical uncertainty. The parameterization approach and data assimilation method both obtains a good history matching result to honor the observation data. The example application for closed-loop production management illustrates that the closed-loop operation strategies effectively decrease the geological uncertainty and provide a reasonable estimate reservoir model, and the optimized well controls are fairly smooth and significantly improve the effect of waterflooding with higher economic benefits than the reactive control strategy, which proves the proposed approaches can be applicable for lagre-scale reservoir closed-loop production optimization.