History Matching of a Realistic Stochastic Fault Model Using Fractal Geometry and the Gradual Deformation Method

Abstract

Subseismic faults, which strongly influence the hydrodynamic behavior of fractured reservoirs, are inadequately handled in most studies. First, simulation methods were developed to generate fractured models respecting geological observations and static data. Then, history matching methods were introduced to make these models consistent with dynamic data too. However, these fracture models were too simple to accurately describe geology. In this work, we go one step further and develop a methodology to calibrate a realistic stochastic fault model to both static and dynamic data. We first propose a fractal methodology to generate the faults. The model parameters (length distribution, clustering properties, and orientation) are derived from the analysis of the seismic faults. A stochastic algorithm is used to randomly draw 3D object based sub-seismic fault realizations. Then, these realizations are constrained to the faults identified from seismic. This results in geologically consistent static fault models. The fracture network is then discretized on a dual porosity simulation grid. Equivalent flow parameters are computed using an analytical method, which makes it possible to account for large-scale fractures. Last, full field simulations are performed using a commercial multiphase flow simulator. Then, we introduce a method to gradually change the locations of faults while preserving geological consistency. The changes in locations are driven from a reduced number of parameters. During the history-matching process, these parameters are gradually perturbed to compel the initial fault model to reproduce as well as possible the hydrodynamic data collected on the field. As the resulting fault displacement is continuous, common optimization algorithms can be used. The potential of the methodology is stressed from a 3D case study. A sub-seismic fault network is generated and constrained to seismic faults. Equivalent parameters are computed on a complex CPG grid. Then, the model is calibrated to water-cut levels and well pressures by varying fault locations.