Automated SVD filtering of time-frequency distribution for enhancing the SNR of microseismic/microquake events

Abstract

Recently, there has been a growing interest in continuous passive recording of passive microseismic experiments during reservoir fluid-injection monitoring, hydraulic-fracture monitoring and fault-movement monitoring, to name a few. The ability to accurately detect and analyze microseismic events generated by these activities is valuable in monitoring them. However, microseismic events usually have very low signal-to-noise ratio (SNR), especially when monitoring sensors (receivers) are located at the surface where coherent and non-coherent noise sources are overwhelming. Therefore, enhancing the SNR of the microseismic event will improve the localization process over the reservoir. In this study, a new method of enhancing the microseismic event is presented which relies on one trace per receiver record unlike other methods. The proposed method relies on a time-frequency representation and noise eliminating process which uses the singular-value decomposition (SVD) technique. Furthermore, the SVD is applied on the matrix representing the time-frequency decomposition of a trace. More importantly, an automated SVD filtering is proposed, so the SVD filtering becomes observation-driven instead of user-defined. Finally, it is shown that the proposed technique gives promising results with very low SNR, making it suitable to locate passive microseismic events even if the sensors are located on the surface.