Conditioning reservoir models to dynamic data is challenging due to the non-linear relationship between the measured flow response data and the model parameters (porosity, permeability etc.). The focus of this paper is to present a methodology for efficiently integrating dynamic production data into reservoir models. In contrast to other methods for production data integration, the proposed methodology attempts to quantify the information in production data pertaining to reservoir heterogeneity in a probabilistic manner. The conditional probability representing the uncertainty in permeability at a location is iteratively updated to account for the additional information contained in the dynamic response data. A localized perturbation procedure is also presented to account for multiple flow regions within the reservoir. The proposed methodology is demonstrated on a realistic case example. The methodology for implementing the proposed algorithm on parallel cpu's is presented. The computational synergies realized through domain decomposition flow simulation are likely to result in significant cpu savings.
Accurate prediction of reservoir performance is necessary for planning and implementing critical reservoir development decisions. Accurate models for spatial variations in rock and fluid properties are necessary for making reliable predictions of future reservoir performance. This in turn predicates efficient synthesis of all available information pertaining to the reservoir. Production data in the form of well bottom hole pressures and fluid production rates are especially informative about spatial variations in reservoir attributes. However, calibrating the information contained in dynamic data pertaining to spatial variations in static reservoir attributes (such as porosity, permeability, etc.) is non-trivial precisely because of the non-linear relationship between such static parameters (spatially varying petrophysical properties) and the observed dynamic response (e.g. well pressure as a function of time). Several geostatistical algorithms are available to condition reservoir models to static data, but to date, few robust algorithms for conditioning models to dynamic data exist.
Number of Pages
Looking for more?
Some of the OnePetro partner societies have developed subject- specific wikis that may help.