Monitoring Co2 Storage At Cranfield Mississippi With Time Lapse Offset Vsp Using Integration And Modeling To Reduce Uncertainty

Author : Ismael Dawuda
Publisher :
ISBN 13 :
Total Pages : 0 pages
Book Rating : 4.:/5 (144 download)

DOWNLOAD NOW!

Download or read book Geologic Modeling and Data Assimilation for CO2 Sequestration in Point Bar Reservoirs written by Ismael Dawuda and published by . This book was released on 2024 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: The target reservoirs in many CO2 sequestration projects exhibit point bar geology characterized by the presence of shale drapes that can act as barriers to prevent the leakage of CO2. However, these shale drapes can also act as flow barriers and impede the displacement of CO2 in such reservoirs and restrict the storage volume. Therefore, developing a framework for modeling point bars and their associated heterogeneities is important. Yet, for the point bar model to be geologically realistic and reliable for predicting the displacement of the CO2 plume during sequestration, it should be calibrated by assimilating historical production/injection data to reduce the uncertainties associated with predictions of flow performance. Even so, due to the complex geologic heterogeneity exhibited by point bars, there is likely to be significant residual uncertainty even after assimilating historical flow performance related data. The calibrated models are further refined by assimilating timelapse seismic data in a Bayesian model selection workflow to sub-select the most-probable models that best reflect the reservoir characteristics closely. Given the interlinked nature of these modeling efforts, this dissertation proposes an integrated modeling workflow to accomplish the research objectives. The workflow begins with detailed geometric and geologic modeling of point bar reservoirs, and subsequent calibration of the models by assimilating CO2 injection data and time-lapse seismic information. A stochastic approach that considers the processes leading to the deposition of the point bar is proposed to model the point bar and its associated heterogeneities. The method uses geometric functions to model the areal and vertical dimensions of the point bar reservoir. Preserving the curvilinear continuity of the point bar geometry is very difficult and this has been accomplished by implementing a gridding scheme that accounts for the aerial geometry of the accretion surfaces as well as the sigmoidal geometry of the inclined heterolithic stratifications. Also, the spatial continuity of the unique heterogeneities that characterize point bar reservoirs was honored by incorporating a grid transformation scheme in the geostatistical simulation of the reservoir properties. The residual uncertainty associated with the geological modeling process was represented by generating several realizations of point bar reservoir models. The model calibration workflow seeks to reduce the uncertainty associated with the prediction of reservoir properties over the ensemble of point bar reservoir models. The workflow developed in this research addresses two challenges common to many history matching techniques: (1) failure to account for uncertainties in reservoir geometry despite the influence that the reservoir architecture can have on reservoir response variables, (2) inability to handle the non-Gaussian relationship between the primary state variables and secondary variables for reservoirs with complex heterogeneities (such as point bars) within current ensemble-based schemes. These challenges were addressed in a hierarchical, two-step approach using ensemble-based data assimilation techniques. In step 1, we tackled the first challenge by implementing ensemble Kalman Filter (EnKF) to update the geometry of the point bar reservoir. For step 2, we used the updated reservoir geometry determined in step 1 to tackle the second problem by implementing a modified Indicator-based Data Assimilation (InDA) to update the permeability distribution in the point bar system. To accommodate the curvilinear geometry of the reservoir implemented while still implementing InDA in a Cartesian framework, we incorporated a grid transformation scheme. This two-step model calibration approach reduces but does not eliminate the uncertainty associated with the models for the point bar reservoir. Further reduction in uncertainty is possible by integrating additional data in the form of time-lapse information. In this research, we implement a Bayesian model selection workflow to further reduce the uncertainty associated with the models for the point bar reservoir. The model selection algorithm is used to create a posterior set of models that reflect the time-lapse seismic information that may be available for the field site. The algorithm proceeds by: (1) computing discrete Fréchet distances to quantify the similarity in post-injection seismic responses obtained from a large prior ensemble of models, (2) combining multidimensional scaling with k-means clustering, to partition the models into subgroups based on their seismic responses, (3) performing Bayesian computations in the reduced model space to select the subgroup of models that yield response closest to the observed seismic information, and (4) iteratively sampling the posterior models, to further refine the selection of the model clusters. The applicability of the entire integrated workflow to a real field scenario is demonstrated, using the CO2 injection and timelapse seismic dataset for the Cranfield reservoir in Mississippi. The final ensemble of selected models can be used to assess the uncertainty in predicting CO2 storage capacity and the future displacement of CO2 plume.