Multiscale Determination of In Situ Stress and Fracture Properties in Reservoirs

Abstract

In this thesis we address the problem of determining in situ stress and fracture properties in reservoirs using borehole logs and surface seismic reflection data. The dissertation covers four subtopics. The first is the determination of horizontal stress magnitudes from measurements in a borehole. Two types of data used are stress-induced rock failures in the borehole, known as "breakouts," and the dispersions of polarized flexural waves which propagate along the borehole. Traditionally these data are analyzed to derive stress orientations but not magnitudes. To determine the magnitude of stresses directly from breakouts, we use an iterative elastic modeling of stresses around the borehole and Mohr-Coulomb failure criterion to match the borehole deformation. As a second method we use dispersion curves of the two polarized flexural waves and their crossover points. These methods are applied to data from a well in northeastern Venezuela. The combination of these two techniques provides a complete profile of stress as a function of depth since the first method is applied at the breakout depths and the second is applied everywhere else in the borehole. Both borehole methods agree in the estimation of stress orientation and magnitude. The maximum horizontal stress is in the NNW-SSE direction, in agreement with a regional stress model calculated from the relative motions of the Caribbean and South America plates. The magnitudes of principal stresses are on average, SHmax ≃ 1.1Sv (Sv: vertical stress) and Shmin ≃ 0.9Sv (Shmin: minimum horizontal stress). This suggests strike-slip faulting, consistent with earthquake mechanisms in the region. The in situ stresses play an important role on determining the properties of fractured formation. The azimuth of SHmax determines the preferred orientation of open fractures. Surface seismic reflection data provide the means for detecting the fractures. The second contribution of this thesis is developing a method to detect discrete fractures, and to determine their orientation and average spacing. We developed a novel and practical technique, called the F-K method, based on the frequency-wavenumber (f-k ) domain analysis of seismic coda. The fractured medium targeted in this study is a network of rather regularly spaced, parallel, sub-vertical fractures, with dimensions similar to seismic wavelength. The seismic response of a fractured medium is studied by finite difference numerical models for a variety of situations where orientation, spacing, height, and fracture compliance are varied. In the direction normal to fractures, scattered waves propagate with slower apparent velocities than waves propagating along the fractures. The orientation of fractures is well constrained from the azimuthal dependence of scattering. The spectral characteristics (frequency, wavenumber and amplitude) of the backscattered waves are related to fracture properties like spacing, compliance, and height. The dominant wavenumber is very sensitive to fracture spacing. We use the F-K method to analyze a data set from the Lynx Field in Canada. Characterization of fracture properties in this field is important for development plans to maximize the gas production. In the field data, the acquisition geometry results in irregular fold, with under sampling of certain azimuths and offsets. We address the acquisition footprint issue by controlling the azimuth binning of the data and neglecting the low/irregular fold gathers in the fracture analysis. We also apply the Scattering Index (SI ) method (Willis et al., 2006) to the same data from the Lynx Field. The SI method is a robust method to detect fractures and to provide fracture orientations using multi-azimuth/multi-offset pre-stack data. In the realm of existing 3D seismic surveys, data with such acquisition characteristics are few. The fourth contribution of this thesis is therefore the conception of a post-stack version of the SI method that extends the scope of this method to practically every 3D seismic surface data set. In this version, a scattering index is computed for a fully stacked trace per CMP gather. As long as the bin contains traces parallel to the fracture strike, the stacking process of all azimuths and offsets preserves the reverberating character introduced by the fractures. The post-stack SI at a fractured location has a large value in comparison to a non-fractured location. The variations of post-stack SI values across the field reveal the distribution of highly fractured areas. Fracture strike cannot be determined in this case because it does not include the azimuthal behavior of the scattering. However, the results from the post-stack SI are helpful to identify areas of interest to focus the more specialized scattering analysis methods. We apply the F-K and SI methods to the Lynx Field seismic data and compare the results. Since spatial resolution of the two methods are different we upscale the SI maps to match the resolution of the F-K method. The combined analysis of the Lynx Field indicates that the preferred fracture orientation is N40°E, which agrees with the regional stress field. The distribution of highly fractured regions appears to be associated to the geological features, such as folds and faults. The average fracture spacing, obtained by the F-K method shows that, in the Lynx Field, fracture spacing decreases in the west side of the field where the structural dips are higher.