AAPG Annual Convention and Exhibition

Datapages, Inc.Print this page

Benchmarking and Calibration of 3-D Geomechanical Models


When the stress state is unknown, regional orientations of the closest or most recent faults are used as a proxy. However in areas of complex faulting with presence of salt domes such assumptions results in misleading interpretations. Knowledge of stress orientations and magnitudes can be beneficial not just for drilling purposes but for considerations when evaluating options of matrix vs fracture injection. It is shown that building a 3D dynamic geomechanical model from only five wells of a deep water field, can give fair predictive results in which caliper analysis of a recently drilled well, shows breakouts nearly 90 degrees from the SHmax orientation calculated in the 3D model, corroborating the accuracy of stress predictions. Furthermore, contrary to what would be expected from geological observations, the SHmax orientation is not aligned with the dip azimuth of the nearby fault but rotated 50 degrees from it. Thus, the trajectory of an injector well can be oriented in a favorable direction requiring lower fracturing pressures for hydraulic fracturing operations and providing preferential sweep matrix injection towards the producer (i.e, minimizing water breakthrough risks). Separately from stress orientation and magnitude validation, shear deformation can be used as an extra parameter for calibration/benchmarking, something that is commonly overlooked in coupled fluid flow geomechanics modeling. By incorporating elasto-plastic rock behavior, the results showed predictions of shear deformation either at initial conditions (i.e., prior to depletion) or during subsequent injection periods and in locations that were also corroborated from 1D models. Another practical implication of this is that models can be calibrated to pre-production strain levels without the effort of restauration models. Thus a calibration processes require the modeler to assess how when the model outputs is affected by the model inputs and/or the input data, making 3D geomechanical modeling a predictive tool, rather than just a matching exercise of past events.