Hybrid Optimization Technique Allows Dynamic Completion Design and Control in Advanced Multilateral Wells with Multiple Types of Flow Control Devices

Abstract

Designing a well completion for multilateral wells with multiple types of flow control devices (FCDs) can be a challenging optimization task due to a large number of correlated control variables and computationally demanding objective functions. Consequently, standard optimization workflows may fail to find the optimal design. The lack of a reliable optimisation workflow has forced the industry to adopt a simplified, snapshot approach to intelligent completion design, ignoring long-term dynamic reservoir performance. In this work, a multistage optimization workflow named hybrid optimization (HO), has been developed for effectively optimizing the completion design of multilateral wells that are equipped with multiple types of FCDs. Differential evolution (DE), a metaheuristic optimisation algorithm, is utilized for initial exploration of the search space to identify promising regions, while the generated data are employed to develop a fast surrogate model to mimic the dynamic performance of the computationally expensive reservoir model. Global sensitivity analysis using the Sobol method is then performed with the aid of the developed and tested surrogate model, to divide control parameters into high and low impact groups. The Simultaneous Perturbation Stochastic Approximation (SPSA) algorithm is employed at the final optimisation stage to perform a refined search in the optimal areas previously identified. The proposed framework offers engineers a set of guidelines to adjust the completion design, by modifying the most critical design parameters, in order to maximize production performance while minimizing installation and operational risks. The new workflow has been tested on a 3-D, synthetic, representative reservoir model developed by an intelligent dual-lateral well equipped with inflow control devices (ICDs) inside the laterals, and interval control valves (ICVs) at the laterals’ junctions. The developed HO technique showed superior performance as compared to the current, standard optimization options relying on a single algorithm. It allows efficient dynamic optimization and delivers reliable results in a reasonable time, to replace the snap-shot designs which can be sub-optimal due to their dependency on a single timestep.