Efficient workflow for optimizing well controls

Due to the large dimensionality of typical rate optimization problems, most approaches for field optimization are local search methods. Achievable optima by these methods are directly impacted by the choice of the initial guess for the sought optimal profiles. Here we propose, first, a new formulation of the waterflooding optimization problem to reduce the number of the optimization variables considerably, and, second, a simple low-cost framework to efficiently initialize local-search optimization algorithms, based on accepted reservoir engineering concepts. The initialization strategy is applied to history matched model of the Brugge field (Peters et al., 2010), prepared for a closed-loop reservoir management comparative study in connection with SPE Applied Technology Workshop in June 2008. Also we evaluate three different optimization algorithms and compare the results with two previous works (Asadollahi and Naevdal, 2010 and Lorentzen et al., 2009). The optimization algorithms are pattern search Hooke–Jeeves, reflection simplex Nelder–Mead and sequential quadratic programming. The optimized variables in terms of the achieved net present value clearly outperformed the best solution obtained so far on these history matched models.The workflow has several advantages. First, it is simple and easy to program in every reservoir simulator. Second, it needs only one single forward simulation to obtain a reasonable initial solution. Third, the nonlinear constraints are handled by the simulator and therefore do not need a complex implementation of the nonlinear inequality constraints. Finally, the workflow is in line with the existing optimization approaches and the output from the workflow can be used as input to the other algorithms for further improvement of the results. Among the optimization algorithms the Hooke–Jeeves method performed better than the other algorithms.

► A guide rate approach for waterflooding optimization.
► Evaluation of three different optimization algorithms.
► Significant reduction of variables by reformulation of optimization problem.