Time-lapse target-oriented crosswell full waveform inversion without downhole sources

Time-lapse target-oriented crosswell full waveform inversion without downhole sources

Minato, S. (TU Delft Applied Geophysics and Petrophysics; OYO Corporation)
Ghose, R. (TU Delft Applied Geophysics and Petrophysics)

2019

Time-lapse seismics has a wide range of application in different scales, from near-surface to resource exploration. Crosshole seismics is used to characterize fluid reservoirs and to obtain highly resolved rock/soil-dynamic parameters e.g., elastic moduli and Poisson ratio. Developments in distributed acoustic sensing shows the potential of deploying permanent downhole receivers at low costs. In order to achieve an efficient and accurate time-lapse seismic measurement in such scenarios, we have developed a nonlinear waveform inversion to reconstruct velocity structure between boreholes using VSP data with source located only at the surface, and no downhole sources. The new approach formulates the forward modelling using wavefield representation theorem, which enables directly estimating the velocity structure by minimizing data residuals and calculating the gradient from the adjoint state problem. We test the approach using numerical modelling of time-lapse VSP data to detect layer-specific temporal changes. A heterogeneous shallow vadose zone represents a low-velocity layer. The results show that the new approach provides more stable and more accurate temporal velocity profiles than conventional full waveform inversion, when the initial velocity model does not include the shallow low-velocity layer. The new approach is robust and highly advantageous as it does not require downhole seismic sources.

http://resolver.tudelft.nl/uuid:38577dea-4583-443b-9cd5-a71a1b3582f6

EAGE

Near Surface Geoscience Conference & Exhibition 2019: 8-12 September 2019, The Hague, Netherlands

Near Surface Geoscience Conference & Exhibition 2019, 2019-09-08 → 2019-09-12, The Hague, Netherlands

Accepted author manuscript

Institutional Repository

conference paper

© 2019 S. Minato, R. Ghose