Reservoir characterization using crosswell electromagnetic inversion: A feasibility study for the Snorre field, North Sea

Abstract

The spatial resolution of a commercially available crosswell electromagnetic (EM) system is demonstrated using models derived from three time steps from a reservoir simulation of the Snorre field in the North Sea. The numerical simulation of the Snorre field waterflood shows that crosswell EM field measurements provide high sensitivity to changes in the reservoir over time. This sensitivity is achieved by combining the reservoir geometry derived from surface 3‐D seismic interpretation, reservoir conductivities at well locations, and constrained EM inversion of the reservoir’s electrical conductivity. Inversions of 2‐D and 3‐D numerical models show that the changes in electrical conductivity attributable to changes in water saturation can be quantitatively mapped as a function of time. The inversions provide smooth estimates of the spatial variation of reservoir electrical conductivity that can discriminate between the level of water saturation at different stages of the waterflood. Inversions performed on 2‐D data show that for the Snorre example, 3%–5% Gaussian random noise (depending on the model) can be added without a significant degradation in the inverse models. Two‐dimensional inversions of the full 3‐D data in the Snorre example can map the vertical average electrical conductivity within the reservoir in the interwell region almost as well as when the model is two dimensional (constant in strike direction). The effect of 3‐D structure does not seriously degrade 2‐D inversion in the Snorre example‐even between wells that lie in a line parallel to structure. A series of 2‐D inversions where various constraints and starting models are used demonstrates the importance of incorporating a priori information in the form of starting models and restricting the inversion domain to the reservoir zone. These tests show that totally unconstrained, smooth inversions of the interwell volume provide very limited quantitative information. However, when the reservoir geometry is constrained by seismic data and starting models are provided by linear interpolation of conductivities at well locations, the reservoir’s vertical average electrical conductivity can be predicted to within a few percent by 2‐D inversion. The snorre field consists of a full‐scale reservoir with interwell spacings that exceed 1 km where previous work has demonstrated the applicability of crosswell EM in shallow reservoirs with well separations on the order of 100 m. The simulations show that, given current transmitter and receiver technology, the magnetic fields could be measured in the Snorre field in steel‐cased wells separated from the transmitter by up to 725 m.