Applications of boundary-preserving seismic tomography for delineating reservoir boundaries and zones of CO2 saturation

Abstract

Delineating reservoir units is still a challenge for seismic approaches. Even high-resolution crosswell tomographic approaches that produce smooth velocity models to match traveltime data usually provide limited information about the boundaries of subsurface targets. A recent development of seismic traveltime tomography incorporated with a boundary-preserving regularization constraint promisingly helps to resolve ambiguities in reservoir boundaries, while allowing lateral variations. We applied a kind of boundarypreserving traveltime tomography to delineate boundaries of the reservoir and CO 2 -saturated zones. We chose the minimum gradient support as the regularization to preserve boundaries of the geologic target by penalizing smaller model gradients and smoothing small model variations caused by noise. We evaluated several synthetic and real data applications. Synthetic examples demonstrated that the boundary-preserving algorithm produced improved recovery of the profile shape and velocity values of the blocky targets. Two real applications were for delineating the top and base of the reservoir in the King Mountain field, and for delineating a CO 2 -saturated zone in the McElroy field. These inversion results suggested that the boundarypreserving inversion is able to provide better delineation of the top and base boundaries of the reservoir and boundaries of the CO 2 -saturated zone than the conventional smooth-constrained inversion.

Visco-acoustic data modeling using optimum layer approximation technique: greenwood oil field example, USA

“…The velocity logs at the source and receiver wells are interpolated over the inter-well space. The interpolated velocity is tomographically inverted using ray-tracing method (Tikhonov and Arseni 1977;Langan et al, 1985;Nocedal and Wright, 1999;Zhu and Harris, 2015). The velocity model obtained after five iterations and the RTM results are shown in Figure 6.…”

“…Modelling of seismic data First, Elastic data were simulated for the full survey using the Finite Difference Method, the geological model shown in Figure 1 and the survey described above. Procedure for the elastic data modelling using Finite Difference Methods are described in many geophysics papers (e.g., Hu et al, 1988;Bube and Langan, 2008;Zhu and Harris, 2015). Next, visco-acoustic data are simulated for the same survey using the visco-acoustic Simulator described above and 3046 geological layers model.…”

Binary recovery via phase field regularization for first-arrival traveltime tomography

“…Unfortunately, the forward problem is not well posed for general binary minimizers due to the discontinuity. A natural regularization in the optimization framework is to penalize the size of the interfaces between constant values [45,54,36] or to use other so-called 'blocky' or 'edge preserving' regularizations [31,32,55]. This perimeter regularization is difficult for binary priors because of nondifferentiabilty unless one tracks the interfaces.…”

“…The simplest choice α = 2 is problematic because s ∈ A ⊂ H 1 (Ω) does not give a well posed forward problem. A natural regularization for a binary recovery problem is to penalize the interfaces between constant values [45], (also [54,55]). If s ∈ BV (Ω, {a, b}), we can interpret (3.1) with α = 1 as the total variation of s, that is, the perimeter of the set {s = a}.…”

First‐arrival traveltime tomography with modified total‐variation regularization

“…Youzwishen and Sacchi () employed this functional norm to estimate piecewise‐constant velocity models from synthetic seismic data. Zhu and Harris () applied the MGS approach to crosswell field data and delineated the boundaries of reservoir and CO 2 ‐saturated zones. The compactness measures are preferred instead of smoothness if the subsurface models are supposed to include sharp interface features.…”