A Direct Overparameterize and Optimize Method for Stratigraphically Consistent Assisted History Matching of Object-Based Geomodels: Algorithm and Field Application

Abstract

Summary Object-based static models are typically constructed for stratigraphically complex reservoirs. In this approach, the stratigraphic architecture is represented by use of distinct objects with specific geometric attributes and petrophysical characteristics. It is a major challenge to condition such models to production data while simultaneously maintaining the geologic realism and static conditioning. A novel work flow is developed for the assisted history matching (AHM) of object-based geomodels where the uncertain object locations and attributes are directly modified without resorting to (post-geomodeling) reparameterization techniques. It contains an object-modeling algorithm for channels and levees that provides direct access to the preraster geomodeling parameters for individual object locations and attributes. These parameters are gradually modified subject to physical constraints to achieve a history match. A fully integrated protocol is used in this process that automatically couples the static modeling algorithm with the reservoir simulator. The resulting work flow is moderated by a massively parallel and highly efficient iterative data-integration algorithm. In the AHM work flow, static and dynamic conditioning operations are respectively driven by separate objective functions and are performed at the iteration level in a sequential fashion. The static-conditioning step may add and remove objects in the geomodel, which changes the number of active AHM parameters over the course of the iterative search. The work flow handles such operations with minimal impact on the robustness of the search. A potential application of the direct AHM work flow for object-based geomodels is the identification of locations and attributes of channels in deepwater turbidite reservoirs, where the channels are typically below the resolution of the seismic data; the well spacing is typically larger than the characteristic object dimension, yet the production data exhibit strong sensitivity to channel connectivity. The concept of gradually adjusting the channel locations with the information in the production data (while maintaining static conditioning) is demonstrated on a real data set for a deepwater channelized-turbidite reservoir. The models proposed by the new AHM work flow not only improve the difficult-to-history-match injected-gas-breakthrough profiles but also provide geologically based explanations for them, taking into account the channel connectivity. The proposed AHM work flow ensures consistency across static and dynamic models by integrating multidisciplinary data with an easily auditable and replicable capability.