Ray-based stochastic inversion of pre-stack seismic data for improved reservoir characterisation

Ray-based stochastic inversion of pre-stack seismic data for improved reservoir characterisation

van der Burg, D.W.

Wapenaar, C.P.A. (promotor)

Civil Engineering and Geosciences

2007-12-03

To estimate rock and pore-fluid properties of oil and gas reservoirs in the subsurface, techniques can be used that invert seismic data. Hereby, the detailed information about the reservoir that is available at well locations, such as the thickness and porosity of individual layers, is extrapolated to all locations in the reservoir on the basis of seismic reflections. Stochastic inversion algorithms also compute uncertainties in the property estimates. Standard inversion techniques invert the seismic reflections, present in the form of band-limited signals called wavelets, from migrated data using a 1D convolutional forward modelling kernel; these methods thereby rely on the preceding migration procedure to take into account the propagation effects of seismic waves travelling through the subsurface. In practice however, inevitably wavelet distortion as a function of reflector dip and reflection angle is present on the migration image, and angle-range substacks for enhancing signal-to-noise ratios blur the reflection-angle information needed for resolving reservoir parameters. Any possible flaws in the migration cannot be accommodated for by the inversion. To overcome the above-mentioned difficulties, in this thesis an alternative approach to stochastic inversion is introduced, in which the original wave path and reflection-angle information is taken inside the inversion; the inversion takes place along the wave paths. This means that the data must be inverted pre-stack before migration, which has the advantages that angle-dependent reflection information is not blurred and that migration-induced wavelet distortion does not occur. The reflection response corresponding with these data is modelled using 3D elastodynamic ray-tracing, which makes it possible to interweave seismic trace-inversion with Kirchhoff-type migration. The new method is called ray-based stochastic inversion, and can be regarded as a generalisation of current amplitude-versus-offset/amplitude-versus-angle (AVO/AVA) inversion techniques. The new method is designed to outperform standard stochastic inversion techniques in cases of reservoir parameter estimation in a structurally complex subsurface with substantial lateral velocity variations and significant reflector dips. Results from synthetic data tests show that in strongly dipping reservoir structures, dip-dependent wavelet stretch due to migration severely deteriorates the reservoir parameter estimates obtained with standard inversion. Ray-based inversion has a much better performance in the cases shown. Finally, in a test on field data from the Gulf of Mexico, a comparison is made between reservoir parameter estimates obtained with a simplification of the new method, the estimates found by conventional stochastic inversion, and the actual values at a well drilled after the inversion was done. Despite the fact that 1D convolutional ray-based stochastic inversion uses only 2% of the pre-stack data, the result indicates it has improved accuracy on the dipping part of the reservoir, where conventional stochastic inversion suffers from wavelet stretch due to migration.

seismic
prestack
Bayesian
inversion
ray tracing
reflection
reservoir

http://resolver.tudelft.nl/uuid:2bff83fb-f233-41a7-be1e-480131535f85

978-90-9022521-0

Institutional Repository

doctoral thesis

(c) 2007 D.W. van der Burg