History Matching and Optimization Using Stochastic Methods: Applications to Chemical Flooding
A typical lifecycle of an oil and gas field is characterized by three stages: primary recovery by natural depletion, secondary recovery by fluid injection, and enhanced oil recovery (EOR). The primary goal of reservoir management is to increase hydrocarbon recovery while reducing capital and operational expenditures. Two key techniques for the success of reservoir management are model calibration and production optimization. History matching is used to calibrate existing geological models against to measured data and predict the range of future recovery. Production optimization on calibrated reservoir models provides economic assessment of different field development plans and suggests optimal strategies to maximize recovery and minimize cost.
We first presented the workflow of history matching in chemical flooding. Evolutionary algorithms are the method of choice due to its capability of calibrating various parameter types and its global search nature. Chemical flooding simulator UTCHEM, developed by The University of Texas at Austin, is coupled during the history matching process to consider complex mechanisms such as phase behavior, chemical and physical transformations, etc.
Next, we implemented the proposed workflow to calibrate models in multiple stages that can efficiently reduce large amounts of uncertain parameters in alkaline-surfactant-polymer (ASP) flooding. Each stage of model calibration will follow an order of field scale, and then individual well scale, with consideration of behaviors brought by ASP flooding, such as surfactant/polymer adsorption. The proposed multi-stage history matching workflow is powerful to deliver better history matching results and significantly reduce the uncertainty of large numbers of parameters involved in chemical flooding.
Lastly, we extended the evolutionary workflows for multi-objective optimization via introducing the concept of Pareto optimality. Pareto front method is proposed to handle conflicting objective functions such as oil production and chemical efficiency instead of weighted sum method in optimizing ASP flooding. Non-dominated Sorting Genetic Algorithm (NSGA-II) is used to search for Pareto optimal solutions.
The robustness and practical feasibility of our approaches have been demonstrated through both synthetic and field examples.