Joint Inversion of Time-Lapse Crosswell Seismic and Production Data for Reservoir Monitoring and Characterization

We present a fluid-flow constrained inversion approach for joint interpretation of time-lapse seismic and production data. In this approach, the full-waveform seismic inversion is integrated into the traditional history-matching process. Hence, the interpretation of time-lapse seismic data is constrained by the fluid-flow physics in the reservoir. The key component in the workflow is a fluid-flow simulator, which computes not only the production data in the wells, but also the temporal and spatial distribution of fluid properties, such as fluid saturation and pressure. These fluid properties, together with prior rock properties, can be transformed to acoustic properties using the prescribed petro-elastic model. A finite-difference frequency-domain acoustic solver is then used to simulate the time-lapse seismic responses on the reservoir. We use a multiplicative-regularized Gauss-Newton scheme to update the reservoir model iteratively until good match between measured and simulated data is achieved. The derivative of seismic data with respect to acoustic properties is calculated using the adjoint method, and then connected to reservoir parameters using a chain rule derived from the petro-elastic model, while the derivative of production data with respect to reservoir parameters is calculated using the gradient-simulator method. A synthetic crosswell example is employed to demonstrate that the estimation of permeability and flooding front movement can be significantly improved from the joint inversion of time-lapse seismic and production data.