Joint inversion of time‐lapse seismic and production data using VFSA with local thermal regulation and pilot point parameterization

Abstract

Summary Estimation of reservoir parameters is im portant for prediction and optimization of oil production. Time-lapse seismic data, commonly acquired in mature oil fields to identify profitable bypassed zones, can be used to estimate reservoir parameters by posing the history matching process as a joint inversion problem. Due to the size of the problem and existence of multiple minima (of the error function), only global optimization techniques, such as Very Fast Simulated Annealing (VFSA) have an opportunity to find the global minimum. However, even for the VFSA technique, the computational cost of the forward modeling (particularly, the reservoir simulation) can make the inversion process impractical in real applications if we consider that many inversions are required to estimate the solution uncertainty. To improve the convergence characteristics of the VFSA technique, we have developed a modification of this technique where a regulation of the local temperature associated with the model variables is driven by the local changes of the error function. We call this modification Local Thermal Regulation (LTR) and show that the time (or number of iterations) spent for convergence can be reduced by 25 to 50%, while maintaining the same degree of ergodicity of the conventional VFSA technique.