Workflows involving reservoir simulation to optimize the drilling, completion, and production of unconventional reservoirs are complex and time intensive due to the many different optimization parameters and the time required to simulate future production. As the number of optimization parameters increases, the number of simulations required for traditional optimization techniques also increases. The number of optimization parameters is effectively limited by the time and cost of the simulation runs required to adequately quantify or define the contribution of each parameter to the optimal solution.
An intelligent sequential sampling (ISS) optimization method was developed that can include many optimization parameters to arrive at an optimum value in significantly fewer computer simulation runs than standard design-of-experiments (DOE) methodologies. The ISS methodology employed is sequentially constructive by using previous simulation results to extract functional relationships and adaptively select an appropriate next simulation case for evaluation; in other words, it learns from previous results and applies its learnings to select a next appropriate sample to evaluate. For comparison, the traditional DOE process identifies a predetermined set and number of simulation cases and then "learns" after all the results are in to determine the optimum parameter values.
To demonstrate the optimization processes, the 5-year, undiscounted payout was chosen as the objective function and applied to a Bakken-based case study. The optimization workflow was tested by optimizing the economics of a greenfield well-pad development area with dimensions of approximately 1 mile by 2 miles. Nine parameters were investigated for their effect on the undiscounted payout of the development area including the total number of wells drilled (well-spacing), initial flowing bottom hole pressure (FBHP), final FBHP, the curvature of the FBHP drawdown, how often the FBHP was adjusted, the drilling sequence of the wells, the time between start-up of the wells, number of hydraulic fracture stages per well, and the hydraulic fracture design size.