其他摘要：The study and interpretation of hydraulically stimulated regions, such as certain unconventional hydrocarbon reservoirs (e.g. Vaca Muerta Formation, Neuquén, Argentina), requires the accurate location of the induced microseismic events. The localization is carried out by means of the analysis of the travel times of the generated compressional and shear seismic waves from the unknown event position to a set of geophones, usually located in a nearby monitoring well. The accuracy of the localization, and thus the characterization of the fracturing process, can be strongly affected by the available seismic velocity model, from which only estimates are known. Also, the underlying medium usually shows an anisotropic behavior, meaning that the velocities of the seismic waves depend on the propagation direction. Therefore, knowledge of the parameters that characterize the anisotropy and an appropriate calibration of the velocities can reduce the errors in the localization of the microseismic events. In this paper we propose a strategy to simultaneously calibrate the velocity model and invert the anisotropy parameters from three-component microseismic data. The strategy relies on the hypothesis that the subsurface is composed of a finite number of horizontal layers with weak anisotropy, a widely used approximation that requires only three anisotropy parameters per layer. The differences between the observed and the calculated travel times, for a known seismic source, are quantified by means of an appropriate objective function that turns out to be non-linear and multimodal. For this reason, we minimize it using very fast simulated annealing (VFSA), a stochastic global optimization algorithm devised to find near-optimal solutions to hard optimization problems. Tests on synthetic data show that the proposed strategy can be used to effectively calibrate the seismic velocities and to provide appropriate estimates of the anisotropy parameters in spite of the severe non-uniqueness of the inverse problem at hand. Also, the stochastic nature of VFSA allows us to obtain the uncertainties of the solutions by repeating the inversion several times. Finally, by means of a simulated microseismic location example, we show the importance of having a well calibrated model to successfully estimate the locations of the hydraulically induced events.
