A mixed integer nonlinear program for oilfield production planning

A multi-period model, the Oil Production Optimization Program, has been developed as a prototype of a tool to assist field operators in maximizing profitability of their producing properties. The model captures both the discrete nature of pump replacement decisions and continuous production rate settings. The solution algorithm uses Generalized Benders Decomposition to separate out the continuous and discrete decision variables. Because it is theoretically possible that the algorithm can give suboptimal solutions, we describe it as a heuristic, but it yields the optimal solution on a small test problem, and intuitively reasonable results on larger, realistic problems for which the optimum is not known in advance. Computation times are just a few minutes on the realistic models, but models with many more wells or periods may take an unacceptably long time. Output and sensitivity findings for one- and three-well cases are discussed. Several planned improvements to both the formulation and algorithm are briefly mentioned.