Understanding proppant transport is critically important in designing effective stimulation systems for low-permeability reservoirs, as it leads to better estimates of the propped fracture dimensions and stimulated reservoir volume. Existing models mostly represent proppant as a continuous fluid phase. This assumption is valid for the conventional fracturing designs, where high viscosity fluid (e.g., cross-linking gels) are used as the carrier fluid. Current fracturing designs mostly use low viscosity fluids (e.g., slick water). As a result, proppants behave more like discrete particles and less like a continuous fluid phase. Existing proppant transport models assume a single planar fracture as the main representation of the geometry of fractures, but the geometry of the subsurface fracture networks is much more complex. In this study I couple computational fluid dynamics with the discrete element method (CFD-DEM) to simulate proppant transport in a complex fracture network. The coupled simulator enables the explicit modeling of the motion of individual particles and offers a more accurate representation of the complex interactions between proppant particles, fracturing fluids, and fracture walls. To calibrate the numerical model, I first conducted validation simulations that imitated a particle settling test, a particle collision test and a laboratory proppant transport experiment. Through scoping calculations, I determined the correct drag force model and matched the model predictions with existing analytical solutions and experimental data for a wide range of flow regimes, including three different sizes of proppants (20-30 mesh, 30-40mesh and 50-70 mesh) in two types of fluids (water and oil). In the main component of my study, I built multiple 3-dimentional fracture network models, which include one baseline vertical fracture model, three dipping fracture models, two hydraulic fracture-natural fracture (HF-NF) intersection models (T-shaped and Z-shaped) and, finally, a multi-cluster horizontal wellbore model. In the baseline vertical fracture model, the simulation results show that the flow regime of proppant (suspension or bedload transport) plays a critical role in determining the proppant advance and distribution in the fracture. Higher fluid velocities lead to a larger suspension transport region and a higher proppant placement efficiency in the hydraulic fractures. In the dipping fracture models, my results show that decreasing the dipping angle increases the proppant placement efficiency. In the T-shaped HF-NF intersection model, I observed significantly better proppant placement in the NF when proppants are in the suspension transport regime. In the Z-shaped HF-NF intersection model, my study identified two parameters that are critical for estimating the occurrence of proppant bridging: the proppant concentration (Cp) and the ratio between the secondary fracture aperture and the proppant diameter (Rfp). At a fixed value of Rfp, continuous transport of proppant is possible when Cp is lower than a threshold value. Based on this determination, I use Rfp and Cp to propose a blocking criterion correlation. Lastly, in my multi-cluster wellbore model, I experimented with various pumping strategies and computed the proppant and fluid distribution at each cluster. By comparing the influence of injection rate, I discussed potential strategies to achieve a better (more even) proppant distribution at the different clusters.
