A Derivative Free Optimization Method for Reservoir Characterization Inverse Problem

Reservoir characterization inverse problem aims at forecasting the production of a field from available production data. These data (pressure, oil/water/gas rates at the wells and 4D seismic data) are compared with simulated data to determine petrophysical properties of the reservoir. The underlying optimization problem requires dedicated techniques : derivatives are often not available, the associated forward problems are CPU time consuming and some constraints may be introduced to handle a priori information. In this paper, we propose a derivative free optimization method based on trust region approach coupled with local quadratic interpolating models of the cost function and of non linear constraints. Results obtained with this method on a synthetic reservoir application with the joint inversion of production data and 4D seismic data are presented. Its performances are compared with a classical sequential quadratic programming method in terms of numbers of simulation of the forward problem.