Abstract
Abstract
While Carbon Capture and Storage (CCS) has a pivotal role in meeting climate change targets, the question remains, "can we adequately predict the CO2 plume dynamics?" The results so far are not encouraging, which raises concerns about the integrity of sequestration projects and needs to be addressed to capitalise on the value of underground storage. This work is focused on developing an adequate understanding of CO2 migration in a storage unit using an ensemble-based 4D Assisted History Matching (AHM) methodology to improve predictive modelling. Additionally, the study will investigate the critical contributing parameters in the spatial and temporal development of the plume.
We perform a sensitivity study for appropriate selection of the compositional model, accounting for relative permeability hysteresis and identifying influential parameters. The high-resolution reservoir simulator coupled with EnHM Ensemble History Matching software developed by TotalEnergies, is employed to integrate static and dynamic parameters in the AHM workflow. We build an ensemble of 100 realisations for facies and petrophysical properties in the initial step using Truncated Gaussian Simulation (TGS) and Sequential Gaussian Simulation (SGS), respectively. These algorithms cater for uncertainties during data assimilation and ensure geological coherency by constraining the models to the prior information.
The correlation between all uncertain model parameters i.e., static and dynamic, and observations is assessed. The ensemble of the models is then modified using correlations to minimise the difference between simulated response and historical data in an iterative manner. The iteration methodology is based on the Ensemble Kalman Filter (EnKF) and is further enriched by considering the time-lapse seismic as an observation dataset. We define 4D signal in workflow by extracting geobodies from seismic anomalies, and the distance to observed geobody and simulated response is treated as objective function.
The proposed methodology of inculcating the AHM workflow with geological uncertainties and dynamic parameters resulted in a good agreement between simulated and field response while respecting geological realism. The workflow addresses the modelling gaps mainly attributed to the lack of iteration between static and dynamic models, and application of fixed multipliers. It calibrates the hundred reservoir models simultaneously and enables us to make robust and reliable predictions. We conclude that the proposed methodology can potentially improve the prediction of plume migration and make well-informed decisions for all stakeholders. Nonetheless, one topic for future work is to convert the simulation model into a petro-elastic model for direct comparison with seismic response to avoid pre-processing for preparation of geobody and improve the rigorousness of the model.