Coupling Numerical Simulation and Pattern Recognition to Model Production and Evaluate Carbon Dioxide Injection in Shale Gas Reservoir

Massive multi-cluster, multi-stage hydraulic fractures have significantly increased the complexity of the flow behavior in shale. This has translated into multiple challenges in the modeling of production from shale wells. Most commonly used numerical techniques for modeling production from shale wells are Explicit Hydraulic Fracture (EHF) and Stimulated Reservoir Volume (SRV). Model setup for the EHF technique is long and laborious and its implementation is computationally expensive, such that it becomes impractical to model beyond a single pad. On the other hand, identifying the extent and conductivity of SRV is a challenging proposition. SRV technique is commonly used to simplify the modeling and the history matching process. In this dissertation, an integrated workflow, which demonstrates a quantitative platform to model shale gas production through capturing the essential characteristics of shale gas reservoirs, is developed. A dual porosity/ compositional simulation model with explicit hydraulic fractures is developed for a pad with six horizontal laterals and 169 clusters of hydraulic fractures in the Marcellus shale reservoir. This pad is history matched using three years of production history. The history-matched model is used to develop Next-generation shale proxy model (data-driven shale proxy model) at the hydraulic fracture cluster level, using pattern recognition technology. Data-driven shale proxy model provides highly accurate simulation results for the methane production in a second, thus making a comprehensive analysis of production from shale a practical and feasible option. The history-matched and depleted Marcellus shale gas reservoir simulation model is used to perform a feasibility study to evaluate CO2 injection process for the purpose of production enhancement and CO2 storage by coupling numerical simulation and pattern recognition capabilities of Artificial Intelligence. Data-driven shale proxy model for CO2 Enhance Gas Recovery and Storage (CO2-EGR&S) is developed, which is capable of accurately replicating the generated injection and production profiles from the numerical simulation model for each cluster/stage and horizontal lateral. Coupled use of the deterministic reservoir model with Data-driven shale proxy model is served as a novel screening and optimization tool in evaluating the viability of residual gas recovery and CO2 storage in depleted (or near-depleted) shale gas formations. It allows running the model in real time and making the uncertainty quantification possible for CO2-EGR&S process