Flow Constrained Reservoir Characterization using Bayesian Inversion

In the context of the EU Joule II, Geoscience II Research programme, a project has been initiated, aimed at improving techniques for reservoir characterization. One of the topics addresses the reduction of uncertainty by conditioning the subsurface parameters to measured production data. A Bayesian setting gives an appropriate formalism for handling the uncertainty reduction. Prior data on the parameters to be estimated are updated using the field measurements and a forward model. This results in an updated (posterior) probability density function for the parameters. From this posterior pdf, which contains all information on parameter correlation and uncertainty, point estimates can be obtained. These point estimates form the solution of the estimation problem and also result in the associated confidence levels. An algorithm has been implemented to perform the Bayesian Inversion formalism. General forward models can be attached. Some examples are given to demonstrate the principles.