Multi-Objective Optimization of CO2 Enhanced Oil Recovery Projects Using a Hybrid Artificial Intelligence Approach

Abstract

Listen

Translate article

Abstract In this paper, a hybrid scheme that couples artificial neural network (ANN) and multi-objective optimizers is structured to co-optimize oil recovery and carbon storage of CO2 - EOR processes. The workflow is developed and validated employing an injection-pattern-based model. A field scale case study is presented to demonstrate the practicability of the workflow. An injection-pattern based reservoir model employing a compositional numerical simulator is established to develop and test the hybrid-optimization workflow. Such a scheme aims at optimizing objective functions including oil recovery factor, CO2 storage and project net present value (NPV). An ANN expert system is trained and employed as a proxy of the high-fidelity model in the optimization process. The ANN model is trained by a robust optimization procedure which is competent to find the best architecture. Particle swarm optimization (PSO) is coupled with the developed proxy model to optimize a weight-aggregated objective function, and multi-objective functions by a Pareto front approach. A field case study is included in this paper. The reservoir model is well-tuned via a rigorous history matching process using the available field data. The aforementioned workflow is deployed to optimize the tertiary recovery stage of the field development. In this paper, the validation results of the proxy model will be compared against results from the high-fidelity numerical models. Investigations focus on comparing the optimum solution found by the aggregative objective function and the solution repository (Pareto front) generated by the multi-objective optimization process. The optimization results provide significant insight to the decision-making process of CO2 - EOR project when multiple objective functions are considered. This study develops a novel hybrid-optimization workflow for CO2 - EOR projects considering multiple objective functions. The robustness of the development is confirmed via a field case study. Moreover, this work investigates the relationship between the solutions of the aggregative objective function and the Pareto front, which provides constraints and reduces uncertainties involved by the multi-objective optimization process.