--> Abstract: Dynamic Characterization of Fracture Geometry through Bayesian Inversion of Production History; #90063 (2007)

Datapages, Inc.Print this page

Dynamic Characterization of Fracture Geometry through Bayesian Inversion of Production History

 

Fetel, Emmanuel1, Sarah Vitel1, Laetitia Mace1 (1) Nancy School of Geology, Vandoeuvre-les-Nancy, France

 

Geometrical parameters (length, orientation, …) of individual fracture may dramatically impact the flow behavior of fracture reservoirs due to the key role of the fracture connectivity. Such parameters are tediously characterized using static data, hence highly uncertain. A successful way to assess them is to consider production history. Classically, history matching approaches use optimization procedures to find only a few best matching models; however, due to the stochastic modeling of DFNs, such methods will usually fail. On the contrary, the approach developed in this paper estimates the multivariate probability density function of parameters representing all the likely models, by accounting for both uncertain geometrical parameters and stochastic fracture locations.

 

First this approach uses an unstructured upscaling procedure based on discrete fractures that faithfully integrates individual fracture hydraulic properties and geometry into a coarse model. Then the relationship between fracture geometrical parameters and production is approximated using response surfaces. In which, the impact of fracture location is treated as a dispersion effect. Finally, bayesian inversion is performed to evaluate the probability of a model to match the production data.

 

The proposed methodology is illustrated on a realistic reservoir model. Five different settings describe the variations of the two considered geometrical variables (fracture length and orientation). For each combination of these variables, fifteen stochastic realizations of DFNs are generated from a single P32 fracture density map. Application results demonstrate not handling stochastic variations in the fracture location mislead the history matching of other fracture geometrical parameters.

 

AAPG Search and Discover Article #90063©2007 AAPG Annual Convention, Long Beach, California