The generation of a reservoir development plan with well locations, given a reservoir property map and a set of infrastructure constraints, represents a very challenging problem. The problem of selecting the optimal vertical well locations was formulated with 3-D information as a MILP problem where the binary variables correspond to the decisions of well locations and vertical well completions. A novel decomposition-based solution procedure is proposed to address real fields characterized by 60,000 grid points. The approach is based on the ideas of refining the feasible set of candidate well locations by applying quality cut constraints. For the solution of the subproblems required at each iteration, a model reformulation was then suggested that enables the efficient determination of the optimal well locations. Two industrial case studies are considered to illustrate the applicability of the proposed procedure to realistic large-scale problems.