No abstract

In this paper, we propose a new methodology for seismic inversion at the reservoir grid scale. The inversion engine adapts the Gradual Deformation Method (GDM) to seismic inversion. GDM is a geostatistical technique, previously used in the field of history matching, that allows to continuously modify reservoir realizations, while preserving spatial variability. The workflow is the following: − Combine the initial petroelastic properties in the reservoir grid with an independent realization, accordingly to the principles of GDM; − Condition the new realization with well logs; − Simulate the synthetic seismic. The misfit between actual and synthetic data is given by the objective function; − Minimize the objective function by perturbing the gradual deformation coefficient; − Assign the optimized realization as the new initial one. This loop is performed until the match with the actual seismic data is satisfactory. We show that the basic GDM formulation gives promising results on a synthetic test case for seismic inversion in a reservoir model.

Time lapse seismic or 4D seismic is a geophysical monitoring tool used by the industry to guide and maximize field development. It has been proven successful in many clastic deposit environments (Gulf of Mexico, Gulf of Guinea, North Sea…) but barely used operationally on carbonate fields. Major concerns are both on carbonate petrophysical sensitivity to pressure and saturation changes, and particularly in the Middle-East on the expected noisy seismic quality. In reality Truth is more complex and some cases emerge showing that good quality 4D seismic in carbonates is possible. The case study presented here belongs to the worst case family: random noise is high level due to very shallow streamer acquisition design during Base survey, and multiples are highly non repeatable and pollute the dataset. However the final result is an added-value and this paper aims to explain that a comprehensive interpretation mixing innovative QC tools and an integrated 2G&R interpretation can allow separating intelligently noise and 4D signal.

No abstract

No abstract