22 novembre 2012 ( dernière mise à jour : 30 août 2016 )
Subject : Quantitative 4D Seismic Imaging in complex media using 2D Full Waveform Inversion
Supervisors : Jean VIRIEUX, Stéphane GARAMBOIS, and François AUDEBERT (TOTAL E&P Pau)
4D monitoring is used in order to assess the evolution of petro-elastic properties of the reservoirs from an initial acquisition (denoted the baseline, often performed before any hydrocarbon production) and so to ameliorate their productiveness. The obtained differential images constitute an important source of information on the evolution of the characteristics of a given field. For an optimum and quantitative interpretation, they must be as correct and precise possible. For monitoring purposes, one of the promising techniques dedicated to assess physical properties changes in target regions is the differential waveform inversion, both in the acoustic and elastic cases.
Full Waveform Inversion (FWI) is a data fitting procedure aiming to develop high resolution quantitative images of the subsurface, through the extraction of the full information content of the seismic data. Beside the exploration application, the FWI method can be also used for monitoring applications, such as oil and gas reservoirs, steam injection, CO2 sequestration, in order to obtain a quantitative image of changes of physical properties in target regions from successive seismic experiments.
The conventional difference method for time-lapse inversion needs to independently invert the two data sets (baseline and monitor sets) and to subtract the final derived monitor model from that of the baseline one in order to obtain a perturbation image of property changes. This procedure might not be so robust because spurious features on both baseline and monitor images could potentially contaminate the differential model.
An alternative strategy consists in inverting only the differential data set to recover a differential image. The differential procedure has been proposed for time-lapse waveform inversion of acoustic data in frequency domain (Watanabe et al., 2004) and inversion of elastic data in time domain (Denli and Huang, 2009). The main advantage of the differential method compared to the difference approach is that the common noise between surveys can be rejected by data differentiation. The final time-lapse image is then more robust to noise contamination.
Dans la même rubrique :