No abstract

We have developed justified, verified and implemented for supercomputers with parallel architecture the reliable instrument for the studying complicated processes of waves' propagation in realistic 3D heterogeneous multiscale models of geological media -the special finite-difference method with local mesh refinement in time and space. On this base, it becomes possible to deal with realistic models and acquisitions and take into account both small-scale heterogeneities of a reservoir and mesoscale variations in overburden. This software was applied for simulation of seismic waves' propagation through realistic synthetic models developed for some East Siberian oil field. We have found that orientation of fracture corridors and fluid saturation of reservoir microstructure has very specific impact in synthetic images of scattered waves, which can be used to develop predictive criteria in real life data processing and interpretation. These criteria are verified by comparison of predictions with well log data (fracture orientation) and permeability (fluid saturation) of a collector by test results.

No abstract

No abstract

Due to the apparatus restrictions the resolution of the micro-CT scans of rock samples and the size of the 3D image are strictly connected. Thus improve in the resolution giving more detailed representation of the rock structure reduce the size of the sample, so that it goes below the representative volume for the estimation of a particular property. In this paper we study four 3D images of different resolution of Bentheimer sandstone. We show that the geometrical and statistical properties of the images (pore-to-core distribution, porosity, specific surface area, topology) can be stably reconstructed from the images of about 100 voxels. However, numerical estimation of the physical properties, such as permeability, requires larger images. To overcome this difficulty we suggest using statistical simulation of the sampels, based on the stably recovered information – image properties. In this paper we considered the simplest method of statistical simulation; i.e. truncated Gaussiam technique, which is based on the information about the total porosity and correlation length of pore-to-core distribution. This approach underestimates the specific surface area of the pore space, but the permeability of the sample was matched accurately.