A reservoir characterization workflow for time-lapse study requires integrating seismic data vintages and well logs information into a single consistent model to delineate changes of reservoir parameters. The research work in this thesis is divided into four parts: 1) pre-stack inverse problems for time-lapse seismic surveys; 2) prediction of shear sonic logs for wells that are not available; 3) a processing flow for optimum seismic repeatability and imaging of the production related change; 4) interpretation of seismic inversions and seismic differences results from time-lapse seismic surveys.
In this thesis, we present three new time-lapse AVO inversion algorithms: 1) total inversion of the differences; 2) inversion of seismic differences only; 3) sequential reflectivity-constrained inversion. The proposed methods were implemented using synthetic data that simulate reservoir conditions at pre- and post-production of a heavy oil reservoir after depletion. The time-lapse AVO inversion schemes simultaneously invert the P-P & P-S seismic data of the baseline and monitor line surveys to estimate the change of elastic impedances and density model parameters. The proposed algorithms have proven their robustness in terms of computation time and stability in the presence of noise.
Predictions of shear-wave logs in wells that do not have dipole sonic are challenging, particularly in heavy oil reservoirs. The linear-regression, robust locally weighted scattering and smoothing (LOWESS), and several other approaches of iteratively re-weighted linear least-squares inversion (IRLS) techniques were implemented to estimate shear-wave sonic logs. The developed computer codes were applied using well logs from three different types of reservoirs (conventional oil, heavy oil, and tight shale oil) in WCSB. The proposed methods guard against outliers, and have shown improvements in predicting shear-wave sonic logs compared to empirical linear relationships.
The time-lapse processing and imaging flow designed for seismic data vintages of Pikes-Peak oil field has overcome many difficulties related to differences in acquisition parameters and seismic noise. The proposed flow managed to track production related changes in the reservoir and clearly imaged induced amplitude changes at the Waseca reservoir channel.
The dimming zones observed in the stack section of the monitor line reveal that the steam pressures have caused fluid move through the top seal into the overlying formation. This finding also correlates with the cloud of amplitude change in colony formation shown in the amplitude differences of the time-lapse stack section. The impedances differences show that the interbedded silty sands provide a pathway for steam movements within the Waseca sand channel. The elastic attributes from the AVO inversion of time-lapse seismic surveys successfully map the hydrocarbon movement at the Waseca reservoir channel.
Conditioning of pre-stack seismic migrated gathers of the Pikes Peak time-lapse surveys remains a challenge and is not covered in this dissertation. Therefore the implementation of proposed time-lapse inversion schemes is left as future research work.