Time-lapse seismic monitoring of producing reservoirs presents strong potential for improving reservoir management. We have implemented a high speed simulation combining streamline fluid flow and ray-Born seismic modeling to examine feasibility and uncertainty of such monitoring. The simulation is applied to an 80% gas-saturated reservoir model with stochastic models of porosity and permeability. We then apply a rapid streamline fluid simulation. The petrophysical, pore pressure and fluid saturation data provide the basis for predicting time-dependent elastic properties of the reservoir using the Gassmann equation. Bulk moduli for the solid grains and fluid properties were chosen to represent reservoir conditions at a depth of 2 km, and dry rock bulk and shear moduli were chosen with the empirical equation from the laboratory data. Given these parameters from the Gassmann equation, a ray-Born algorithm provides a high speed, 3-D simulation of seismic data. We find that the seismic amplitude change before and after water injection is large enough to be seismically visible in the high porosity regions of the reservoir, but the change in low porosity regions is small and may be masked if the signal-to-noise ratio is small. The pore pressure change is also a critical reservoir property that influences the seismic amplitude. Uncertainty of the dry bulk modulus has less influence in high porosity regions than in low porosity regions.
Time-lapse seismic surveys have been successfully applied to monitor fluid changes in a reservoir during production1,2. The acoustic properties in the reservoir is affected by the fluid changes since the bulk density and the bulk modulus of rock change as the pore fluid is replaced. Forward modeling of time-lapse seismic monitoring has been shown to be a quite useful tool for reservoir management3. Because the potential errors or uncertainties in model parameters might affect the seismic amplitude change in time-lapse seismic modeling, it is valuable to test many different cases using an efficient tool for simulating the data. The models used for this purpose are generally 3D and represent comparatively large volumes in the subsurface, which means that both fluid flow and seismic simulations are numerically expensive.
We solve this problem by combining two high speed simulation methods, a streamline method for modeling fluid flow4 and a ray-Born method for synthetic seismograms5. Both methods are applicable to general models, including anisotropic permeability and elasticity. The Gassmann equation provides the link between these two algorithms by computing the elastic properties of the reservoir rocks under varying saturation conditions as a function of time. The resulting joint simulation is very fast, requiring, for example, about one minute to simulate 365 days of fluid flow in a model with 40,000 grid blocks using a SGI Power Challenge computer. The seismic modeling for 10,000 sources and receiver pairs, assuming a homogeneous overburden, requires about 3 hours in an SGI Indigo2 workstation. Therefore, we can readily consider a number of different model conditions in a short time.