Proxy model-driven optimization of CO2 operating condition and hydraulic fracturing design for maximizing EGR-CCS performance in the Duvernay shale formation, Canada

Abstract

Shale gas is a prominent unconventional resource because of the advances in horizontal drilling and hydraulic fracturing, especially in North America. Shale gas reservoirs have also been considered for carbon capture and storage (CCS) to help mitigate CO₂ emissions and allow additional gas production. Enhanced gas recovery with CCS (EGR-CCS) injects CO₂ to displace methane (CH₄), leveraging the higher adsorption capacity of CO₂ in shale. On the other hand, cumulative CH₄ production and CO₂ stored amount depend heavily on the timing of the injection, while economic factors such as natural gas prices and CO₂ tax credits also play a role—often overlooked in previous studies. This study developed a machine learning-based proxy model to predict the net present value (NPV) of an EGR-CCS project by integrating the CO₂ operating conditions, hydraulic fracturing designs, and economic factors. Based on data from the Duvernay shale reservoir, 200 scenarios were simulated to generate a training dataset. Five regression algorithms were tested. Of these five, extreme gradient boosting (XGB) yielded the highest accuracy, predicting CO₂ stored amount, cumulative CH₄ production, and NPV with R² > 0.9. A complete factorial design was then implemented to optimize the EGR-CCS process under varying economic conditions. This comprehensive framework aids decision-making to maximize the economics of EGR-CCS, highlighting the potential of the Duvernay shale formation as a geological carbon storage target.