Summary
Locating microseismic events in a reservoir monitoring system has attracted considerable interest recently due to its ability to image the induced fracture geometry that result from fluid injection and fracturing processes. Time reversed imaging (TRI) techniques are well recognized localization tools developed in the last few years. This technique back-propagates the received full waveforms to focus at its real source location without picking different arrival phases. However, the time reversed images are often contaminated due to the strong noise and other interference in surface microseismic measurements, leading to unreliable location estimation. To minimize the interference of strong noise, we present a multi-scale TRI technique using the shift-invariant dual-tree complex wavelet transform (DTCWT) to decompose the original waveforms into multiple time-frequency domains (different levels). The TRI are then applied to the waveform component at each level. Images of effective components substantially improved the quality of the final image of subsurface microseismic events with a much sharper focus. On the other hand, the images of noise components may reveal the velocity structure which is helpful in event location. In addition, the noise components recognized by the multi-scale TRI can be applied to estimate the background noise level.
Introduction
The geometrical information of induced fractures can be obtained by accurately locating subsurface microseismic events with time reversed imaging (Gajewski and Tessmer, 2005). This technique is based on source-receiver reciprocity and time invariance of the wave equation in non-dissipative media (Larmat et al., 2006), propagating the emitted seismic energy back to its origin without the need of picking arrival time. However, the surface microseismic measurements often have extremely low S/N ratio, making it difficult to identify any effective events. The high-amplitude ambient noises severely degrade the quality of location images, resulting in unreliable source positions.
In this paper, we first introduce a technique employing the shift-invariant dual-tree complex wavelet transform (DTCWT) described by Kingsbury (2001) to decompose the original waveforms into multiple time-frequency domains for the purpose of effective identification of true micro-seismic events. The time reversed imaging (TRI) is then applied to the wavelet component to generate the location image at each scale. The final location image is given by the multiplication of several effective images. Finally, a 3D synthetic data example is presented to verify the proposed technique.
