Data-Driven Multi-Objective Optimization of CO2 Injection for Enhanced Gas Recovery and Geological Storage
Paper ID : 1058-UGESR
Authors
Abolfazl Ebrahimy *1, Jamshid moghadasi2
1student
2Professor of Petroleum Engineering, Petroleum University of Technology
Abstract
CO2 injection into depleted gas reservoirs can simultaneously enhance gas recovery (EGR) and provide geological storage, but the two objectives compete due to early CO2 breakthrough. This study develops a reproducible data-driven workflow for optimizing CO2 injection strategy using a synthetic depleted-gas reservoir model governed by the Peng–Robinson EOS. A Latin-hypercube dataset of 500 simulations was generated based on four decision variables: injection rate, injection BHP, production BHP, and injection duration. Gradient-boosting regression outperformed a neural network in predicting recovery factor, net CO2 storage, breakthrough time, and peak pressure. NSGA II was applied to maximize recovery and storage under a fracture-pressure constraint, and all Pareto-optimal solutions were re-evaluated using the reservoir model. The compromise solution achieved 54.8% recovery and 12.6 Mt CO2 storage, compared with 50.4% recovery and zero storage for depletion-only production. The workflow highlights the importance of proxy verification and provides a foundation for future application to real reservoirs
Keywords
CO2-enhanced gas recovery; geological CO2 storage; surrogate model; NSGA-II; Pareto front; depleted gas reservoir
Status: Accepted (Oral Presentation)