Time-lapse inversion of anisotropic velocity linked to geomechanics
by Yinbin Ma
- Thesis pdf
Table of contents
- Chapter 1: Introduction
- Chapter 2: Connecting geomechanics to seismic model
- Chapter 3: Time-lapse parameter estimation in anisotropic media
- Chapter 4: Practical time-lapse full-waveform Inversion
- Chapter 5: Field data application
- Appendix A: FWI gradient based on the adjoint-state method
- Appendix B: HPC solutions for the large-scale inverse problem
- Appendix C: Notations on the strain, stress, and the stiffness tensor
Seismic reservoirs undergo production-induced changes, such as fluid saturation, realignment of the stress field, and variations in pressure. Repeated seismic surveys, conducted before and after production, can provide a wealth of information about those changes. Traditionally, the analysis of these surveys is limited to seismic attributes, such as reflectivity, impedance, etc, from the backscattered energy on reflections. In this study, we show how to retrieve reservoir properties directly from the seismic data by analyzing the transmission effect on reflections. We use fully-coupled flow and geomechanical simulations to invert for those properties in an integrated approach.
The methodology relies on a time-lapse application of the full-waveform inversion (FWI) algorithm. By linking geomechanical models to seismic properties, we can build an initial anisotropic seismic model that represents the study area. Production-induced changes in the seismic velocity and anisotropy parameters are estimated using third-order elasticity theory. From numerical studies of the model, we conclude that velocity changes can be estimated from the seismic data with limited offset (less than 5 km), but changes in anisotropic parameters are only sufficiently constrained with long offset data. We then apply the time-lapse FWI approach to a 3D dataset acquired at the Genesis field in the Gulf of Mexico, using a total-variation regularization on the model differences. To address the computational challenges this poses, we use large-scale cloud computing and manage to finish time-lapse FWI on a 3D model within one day. A velocity decrease is recovered in the overburden above the reservoir, extending from the reservoir to the ocean bottom. This phenomenon can be explained by production-induced reservoir compaction and overburden dilation. The agreement of the numerical results from the seismic data and the geomechanical modeling verifies the validity of our method.