Formation permeability is one of the most important parameters for reservoir modeling. There is a high degree of uncertainty associated with the estimation of permeability using the conventional history matching, which adjusts the model to only fit the production data. In this paper, we present an approach to improve the estimation of permeability by history matching time-lapse crosswell electromagnetic and production data simultaneously. A multiphase fluid-flow simulator is used to calculate the time- dependent bottomhole pressure at the wells as well as the temporal and spatial distributions of water saturation and salt concentration in the reservoir. The latter ones are transformed into the formation resistivity using a resistivity-saturation formula. A 3D finite- difference electromagnetic solver is used to simulate the crosswell electromagnetic data. A regularized Gauss-Newton approach is then used to update the permeability in an iterative fashion until achieving a good match between the simulated and the measured data. In the inversion process, the derivatives of production data with respect to permeability are computed using the gradient simulator method, and the derivatives of electromagnetic data with respect to permeability are computed using the adjoint method and the chain rule.
展开▼