No abstract

No abstract

Summary Early thermal breakthrough in enhanced geothermal systems (EGS) due to the presence of preferential flow channels is a major challenge that endangers efficient and economic heat extraction in such systems. Previous studies mainly focused on adjusting circulation rates of the working fluid, which still leaves significant amounts of untapped heat behind. Currently, there is a lack of technologies for altering flow distribution within the fracture network to achieve uniform heat sweeping in the reservoir. This work presents a novel concept for making proppants to autonomously control fracture conductivity based on the surrounding temperature. Here, proppants with negative thermal expansion coefficients have demonstrated the capability for appropriate fracture conductivity adjustment as a function of temperature to achieve uniform flow across the fracture network. Particle-particle interactions governing such functions are explicitly modeled, and then the Lattice Boltzmann methods (LBM) is used to determine the potential impact of closure stress and temperature changes on the permeability of the proposed proppant packs. Microscale analyses are further used to determine the required material properties to achieve a certain improvement in the permeability of the proppant pack. Our analyses show an enhancement in permeability and the associated fracture conductivity by half of their initial values. Field-scale analysis further confirms the effectiveness of the proposed concept as 31.4% more heat can be extracted from EGS over 50 years of production when the proposed proppants are used. Such novel proppants may effectively delay thermal breakthrough, sweep heat from larger rock volumes, and elongate the life span of the EGS project.

Due to inherent heterogeneity of geomaterials, rock mechanics involved with extensive lab experiments and empirical correlations that often lack enough accuracy needed for many engineering problems. Machine learning has several characters that makes it an attractive choice to reduce number of required experiments or develop more effective correlations. The timeliness of this effort is supported by several recent technological advances. Machine learning, data analytics, and data management have expanded rapidly in many commercial sectors, providing an array of resources that can be leveraged for subsurface applications. In the last 15 years, deep learning in the form of deep neural networks, has been used very effectively in diverse applications, such as computer vision, seismic inversion, and natural language processing. Despite the remarkable success in these and related areas, deep learning has not yet been widely used in the field of scientific computing specially when it comes to subsurface applications due to the lack of large amount of data to train algorithms. In this paper, we review such efforts and try to envision future game-changing advances that may impact this field.

Depletion of oil and gas reservoirs during production from fractured wells, or fluid injection through frack-packs may cause significant changes in the effective stress that proppants experience inside the fractures. Previous studies have shown that closure stress may change absolute permeability, however, deriving this relationship requires time-consuming and costly experiments that need to be repeated with the change of proppant or field parameters. Here, we propose a methodology to estimate absolute permeability magnitudes of the propped fracture direction by just using limited knowledge about grains' geometry. In this study, Lattice Boltzmann Methods (LBM) is used to simulate fluid flow to resemble flow in propped fractures under reservoir conditions. Due to the large number of particle and complexity of the sitting of proppant particles next to each other, intensive computational power is required to carry out simulations. Therefore, a parallelized LBM code is utilized to accelerate computations. The three-dimensional granular geometry was constructed from two-dimensional micro-Computed Tomography images. The geometry was then used as an input for the finite element analysis to simulate grain deformation/sliding upon applying different stresses on the samples. The captured deformation of the grains' geometry was then exported to the LBM code for calculation of permeability. Upon increasing confining stress on the samples, the proppant particles began to deform, slide on each other and the pore network evolved. While the standard API cell measures only conductivity changes under different normal stresses, the presented method gives a greater advantage by allowing determination of the simultaneous effect of normal and shear stresses on the components of the permeability of propped fractures. It is believed that parts of complex fracture networks may experience some shear stresses, but these effects could not be measured in the lab easily. Preserving core samples with the same conditions for a long time is not feasible. Therefore, using LBM, we present a fast and reliable method to measure absolute permeability of proppants during flowback and production using limited data in the form of micro-CT images or 3D images of the grains forming proppant packs.