Comparison of deterministic with stochastic fracture models in water-flooding numerical simulations

Determination of multiphase flow properties considering the variation of fracture patterns (i.e., number of fracture sets, their orientation, length distribution, spacing, and in-situ aperture) remains a key challenge in reservoirs. In reservoir engineering, one way is by studying outcrop analogs with comparable petrophysical properties and a similar geological history, and incorporating these data into model building, discretization, and numerical simulation. The limitation of directly incorporating attributes measured on outcrops is that this method is error prone because of postburial processes. Mineralized fracture (vein) attributes are good candidates to use as analogs for open fractures formed under in-situ conditions, to establish the relationship between fracture length and aperture and help to reveal the conditions at the time of their formation, and to quantify fracture-induced porosity in rock masses. Vein attributes determined from scan lines and window samples were combined to condition the stochastic generation of fractures using the discrete fracture network code FracMan. Comparison of water breakthrough time and oil saturation at breakthrough was then determined by applying a constant pressure gradient for each realization to simulate water-flooding numerical simulation using the combined finite element–finite volume method. The different stochastic realizations were compared with discrete fracture and matrix models, and we show how the uncertainty in these fracture attributes affects multiphase flow behavior in naturally fractured rocks. Uncertainty in quantifying these attributes has a profound impact for predicting the oil recovery and water breakthrough time based on limited information from boreholes.