A Generalized Varying Oil Compositions and Relative Permeability Screening Tool for Continuous Carbon Dioxide Injection in Naturally Fractured Reservoirs

Carbon Dioxide (CO₂) injection is a viable technique with huge potential if employed efficiently in naturally fractured reservoirs. However, extensive computational resources are needed to evaluate and consider candidates for continuous CO₂ injection projects. In this study, a new workflow is developed that captures universal fluid compositions and relative permeability data for fields undergoing continuous CO₂ injection in naturally fractured reservoirs.

Injection design schemes and parameters for continuous CO₂ for hypothetical reservoir types are constructed using a miscible, dual porosity, compositional reservoir simulation model. An algorithm is developed to capture varying fluid compositions and relative permeability data for the data set used in this study. Data collected from reservoir simulation cases are used to construct two Artificial Neural Network (ANN) based proxy models. Namely, a proxy for performance prediction that utilizes specific injection design parameters to predict reservoir performance and an inverse proxy for injection design that provides injection design scheme for a desired reservoir performance.

The constructed ANN’s predictability is assessed through its capability of predicting blind test data within a specified tolerance. The proxy based models were capable of predicting, within an acceptable degree of error, the oil and gas production profiles as well as injection design parameters of well spacing, CO₂ injection rate, and CO₂ injection duration of a field undergoing continuous CO₂ injection. The constructed networks were tested with real data from Twofreds field in Texas that has been undergoing continuous CO₂ injection since 1974. The constructed proxy based models were able to overcome limitations imposed by the complexity of handling different compositional reservoir fluids and relative permeability data. The new methodology sets a new benchmark for universal proxy modeling as it incorporates composition based reservoir fluid and relative permeability data.

The approach presented overcome existing limitations in handling compositional reservoir fluid variations as well as relative permeability in universal workflows.