Hierarchical stochastic modeling and optimization for petroleum field development under geological uncertainty

• We propose a new hierarchical modeling approach based on principal feature analysis.
• Develop a new retrospective optimization algorithm with hierarchical sampling.
• We apply hierarchical modeling for complex systems uncertainty quantification.
• The new computational approach is used to model and solve petroleum field problems.

Real-world simulation optimization (SO) problems entail complex system modeling and expensive stochastic simulation. Existing SO algorithms may not be applicable for such SO problems because they often evaluate a large number of solutions with many simulation calls. We propose an integrated solution method for practical SO problems based on a hierarchical stochastic modeling and optimization (HSMO) approach. This method models and optimizes the studied system at increasing levels of accuracy by hierarchical sampling with a selected set of principal parameters. We demonstrate the efficiency of HSMO using the example problem of Brugge oil field development under geological uncertainty.