ABSTRACT: Permeability evolution is one of the important phenomena that occurs during injection/production in reservoirs. The separation of permeabilities measured during the loading/unloading process known as permeability hysteresis and is more paramount in heterogeneous formations, such as the Bakken and Three Forks Formations in North Dakota. Several experimental results showed that the permeability follows an exponential trend with respect to effective stress. However, these correlation coefficients are taken from core samples at a certain depth and cannot represent permeability evolution of the entire formation. This paper examines the permeability-pressure relationship dominant in tight fractured formations by utilizing machine learning (ML) approach. An artificial neural network (ANN) model was trained based on the variation of core samples' permeability for a wide range of depths to define a general model and predict permeability alteration as a function of the effective stress changes. The effect of reservoir compaction and permeability damage presented in this work was used to evaluate different CO2 injection scenarios for the Middle Bakken and Three Forks Formation. The results demonstrated that CO2 injection in these formations is a strong function of fracture/matrix permeability damage. As a result, considering permeability hysteresis in numerical simulation can help to understand the role of different injection scenarios and enhancing the knowledge for controlling and managing reservoir production by proper operation decisions in unconventional reservoirs.

This content is only available via PDF.
You can access this article if you purchase or spend a download.