Conditioning 3D Object-Based Models to a Large Number of Data

Abstract

Object-based stochastic simulation models are commonly applied for generating facies or rock models with more realistic representation of complicated reservoir heterogeneity. A limitation of object-based modeling is the difficulty of conditioning to dense well data. One method to achieve conditioning is by applying optimization techniques. The optimization algorithm accesses an objective function measuring the conditioning level of objects and the geologic realism. The objective function is optimized with implicit filtering which can consider constraints on object parameters. Thousands of locally optimized objects are generated and stored in a database. A set of the objects are selected with linear integer programing and populated into the model to honor all well data, proportions and other geologic features. To illustrate the conditioning of object-based modeling with complicated geologic features, objects from fluvial reservoirs are used, although any parameterizable object can be considered with the proposed methodology. Channels, levees, crevasse splays and oxbow lakes are parameterized based on location, path, orientation and profile shapes. Functions mimicking natural river meandering are used for the centerline model. Channel stacking pattern constraints are also included to enhance the geological realism of object interactions. Correlations between different types of objects are modeled as well. Case studies considering the geology of different styles of reservoirs demonstrate the flexibility of the methodology. The reservoir styles modeled include bifurcating channels, braided channels, fragmented channels and reservoirs with four types of objects. In all cases the proposed method robustly adheres to realistic feature geometries and matches the dense well constraints. The methodology can be easily modified and applied to many different geologic settings.