Читать книгу Geophysical Monitoring for Geologic Carbon Storage - Группа авторов - Страница 31

For the first phase of this work, Lawrence Berkeley National Laboratory contracted TRE to process existing data in the European Space Agency Envisat archive from 12 July 2003 to 19 March 2007. Two satellite tracks covered the region containing the three injection wells (Tracks 65 and 294). During this initial study period, the respective tracks contained 26 and 18 satellite images, with one or more months between each image. The data were processed using the permanent scatterer algorithm described above (section 2.2.3). The range change data for both tracks provided an estimate of ground displacement along the line of sight of the satellite, that is, along the look direction of the satellite. The analysis of the Envisat data revealed observable surface deformation associated with the injection of carbon dioxide. Peak velocities of over 5 mm/year were found for both tracks, exceeding the estimated errors of 1 mm/year. Elongated patterns of range decrease were imaged, suggesting uplift over the three injection wells KB‐501, KB‐502, and KB‐503 (Fig. 2.3).

Initially, the range decreases in the region overlying the well KB‐501 were assumed to be the result of injection‐related volume change within the reservoir (Vasco et al., 2008). Following the procedure described above (see equation 2.11) the reservoir volume surrounding the injector was mapped into a grid of cells. A regularized least‐squares approach was used to estimate the fractional volume changes within each grid block. A mapping of the sequence of range change into reservoir‐volume change in the region surrounding well KB‐501 indicated preferential migration to the northwest of the injection well. Using the diffusive imaging technique described by equations (2.18, 2.19, and 2.20). the propagation times of the volume changes were used to calculate permeability variations within the reservoir layer (Vasco et al., 2008).

Figure 2.3 Range changes above the carbon storage site at In Salah, Algeria, 1,261 days after the start of injection. The black lines are the traces of the injection wells within the target formation. The open circles denote the wellhead locations of the gas producers.

Specifically, the time series of volume changes for each grid block were used to define the onset of the peak rate of change T _peak . This time is related to the phase of the propagating front σ , according to (Vasco et al., 2000). In Figure 2.4, the variations in the phase are plotted for those regions where the volume changes were significant. The trajectories are found by solving equation (2.19). The solution is quite simple because the right‐hand side is the known phase field σ . So the trajectories are found by marching down the gradient from each grid block center back to the well. Given the trajectories and the spatial distribution of σ , one can estimate the diffusivity by solving equation (2.20) for κ . This is a tomographic problem of a slightly different nature than that of cross‐well imaging. For the problem at hand, there are many trajectories traveling along a similar flow path but for varying distances. The density of trajectories sampling the flow path depends upon the temporal sampling provided by the geodetic data. Thus, one can incrementally build up the diffusivity distribution along each flow path, based upon the set of trajectories sampling that path. The resulting permeability estimates indicated a narrow, northwest‐trending corridor of higher permeability (Fig. 2.4). The correlation of this high permeability feature with a northwest trending break in seismic topography suggested that the carbon dioxide injected into well KB‐501 might flow preferentially along a fault zone on the flank of the anticline defining the field (Vasco et al., 2008).

The range changes observed over injector KB‐502 displayed a distinct double‐lobed pattern seen in Figures 2.3 and 2.5 (Vasco et al., 2010). As noted by collaborators from Pinnacle Technologies, such patterns were often observed in tilt‐meter monitoring of hydrofracturing, suggesting the tensile opening of a steeply dipping planar feature (Davis, 1983). For a tensile feature, such as a fracture, the fluid pressure induces the fracture to open. Thus, the volume change is primarily due to displacement normal to the fracture plane, a change in fracture aperture (fracture width). On the basis of Pinnacle's suggestion, a vertical to subvertical fault/fracture zone model was constructed. The zone extended 80 m above and below the reservoir and approximately 6 km to the northwest and 6 km to the southeast of the injection point in the plane, trending on an azimuth of 135 degrees. In addition, volume change within the reservoir layer was also accounted for in the fashion described above (Vasco et al., 2010).

Figure 2.4 (a) Onset time of the most rapid change in reservoir volume. The curves indicate the flow paths from the well to points within the reservoir. (b) Logarithm of the permeability multiplier found by solving equation 2.20 along the trajectories.

Figure 2.5 (a) Detailed view of range change above well KB‐502 following 1,060 days of injection. The darker colors signify range decrease associated with uplift above the well. (b) Seismic horizon displaying pushdown, most likely due to velocity decreases associated with the migration of injected carbon dioxide.

A homogeneous elastic model of the overburden did not produce the correct range change estimates, and hence an inversion based upon a model with uniform properties led to erroneous depth estimates for a tensile source. Thus, elastic layering, derived from well logs, had to be included in the modeling. The total range change from 12 July 2003 through 19 March 2007 was used to infer the distribution of volume change within the reservoir and the distribution of aperture change over the fault/fracture zone. It was found that the range changes could be matched by a combination of reservoir volume change and cumulative tensile opening of a fault/fracture zone confined to lie at depths within 100 m of the reservoir (Vasco et al., 2010). The lateral extent of the fracture opening was much greater, extending over 3 km from the injection point. Aspects of the fault/fracture zone were subsequently supported in an analysis of three‐dimensional surface seismic data. In particular, Gibson‐Poole and Raikes (2010) noted that time shifts, thought to be due to the injection of carbon dioxide and the resulting seismic velocity change, followed a remarkably linear zone with near parallel boundaries located between the two lobes of range change (Fig. 2.5). The orientation of the seismic feature agreed rather well with the azimuth of 135 degrees required to fit the range‐change data. These conclusions were subsequently supported by the work of Zhang et al. (2015, 2016).