# Differences

This shows you the differences between two versions of the page.

sep:research:theses:sep146 [2012/01/05 14:49]
gayeni created
sep:research:theses:sep146 [2015/05/27 02:06] (current)
Line 2: Line 2:
//by [[http://sepwww.stanford.edu/sep/gayeni|Gboyega Ayeni]]// //by [[http://sepwww.stanford.edu/sep/gayeni|Gboyega Ayeni]]//
-Thesis ([[http://sepwww.stanford.edu/data/media/public/docs/sep146/sep146.pdf|PDF]])+Full thesis ([[http://sepwww.stanford.edu/data/media/public/docs/sep146/sep146.pdf|PDF]])
-  * Chapter 1: Introduction +  * Chapter 1: [[http://sepwww.stanford.edu/data/media/public/docs/sep146/chap1.pdf|Introduction]]
-  * Chapter 2: Time-lapse cross-equalization +  * Chapter 2: [[http://sepwww.stanford.edu/data/media/public/docs/sep146/chap2.pdf|Time-lapse cross-equalization]]
-  * Chapter 3: Joint least-squares migration/inversion: theory +  * Chapter 3: [[http://sepwww.stanford.edu/data/media/public/docs/sep146/chap3.pdf|Joint least-squares migration/inversion: theory]]
-  * Chapter 4: Synthetic examples +  * Chapter 4: [[http://sepwww.stanford.edu/data/media/public/docs/sep146/chap4.pdf|Synthetic examples]]
-  * Chapter 5: 2D field data application  +  * Chapter 5: [[http://sepwww.stanford.edu/data/media/public/docs/sep146/chap5.pdf|2D field data application]]
-  * Chapter 6: 3D field data application  +  * Chapter 6: [[http://sepwww.stanford.edu/data/media/public/docs/sep146/chap6.pdf|3D field data application]]
-  * Chapter 7: Conclusions +  * Chapter 7: [[http://sepwww.stanford.edu/data/media/public/docs/sep146/chap7.pdf|Conclusions]]
-  * Appendix A:  +  * [[http://sepwww.stanford.edu/data/media/public/docs/sep146/sep146.pdf|Appendices]]
-  * Appendix B: +
**Abstract** **Abstract**
+This dissertation presents methods that overcome some limitations in the application
+of time-lapse seismic imaging to subsurface reservoir monitoring. These methods
+attenuate artifacts and distortions in time-lapse seismic images that are caused by
+differences in survey acquisition geometries, presence of obstructions, complex overburden
+and man-made noise. Unless these artifacts are attenuated, it is impossible
+to make reliable deductions about changes in subsurface reservoir properties from
+time-lapse seismic images.
+Improvements to two conventional post-imaging seismic cross-equalization methods
+are considered. Multidimensional warping of baseline and monitor images is implemented
+as sequential one-dimensional cross-correlations and interpolations. This
+method avoids the cost of full three-dimensional warping, and it avoids errors caused
+by considering only vertical apparent displacements between images. After warping,
+matched filters are derived using optimal parameters derived using an Evolutionary
+Programming algorithm. Applications to four North Sea data sets show that a
+combination of these two methods provides an efficient and robust cross-equalization
+scheme. Importantly, the warping method is a key preprocessing tool for linearized
+joint inversion.
+Linearized joint inversion of time-lapse data sets is an extension of least-squares
+migration/inversion of seismic data sets. Linearized inversion improves both structural
+and amplitude information in seismic images. Joint inversion allows incorporation
+spatial and temporal regularizations/constraints, which stabilize the inversion
+and ensure that results are geologically plausible. Implementations of regularized
+joint inversion in both the data-domain and image-domain are considered. Joint
+data-domain inversion minimizes a global least-squares objective function, whereas
+joint image-domain inversion utilizes combinations of target-oriented approximations
+of the Hessian of the least-squares objective function. Applications to synthetic data
+sets show that, compared to migration or separate inversion, linearized joint inversion
+provides time-lapse seismic images that are less sensitive to geometry differences
+between surveys and to the overburden complexity. An important advantage of an
+image-domain inversion is that it can be solved efficiently for a small target around
+the reservoir.
+
+Joint image-domain inversion requires careful preprocessing to ensure that the
+data contain only primary reflections, and that the migrated images from different
+vintages are aligned. The importances of various preprocessing steps are demonstrated
+using two-dimensional time-lapse data subsets from the Norne field. Applications
+of regularized image-domain joint inversion to the Valhall Life-of-Field Seismic
+(LoFS) data sets show that it provides improved time-lapse images compared to
+migration. These applications show that regularized joint image-domain inversion
+attenuates obstruction artifacts in time-lapse seismic images and that it can be used
+to image several data sets simultaneously. Furthermore, because it is computationally
+efficient, joint image-domain inversion can be repeated quickly using various a priori
+information.