Evaluation Of The Potential For Geomechanical Monitoring And Stochastic Calibration Methods To Improve Characterization During Geologic Carbon Storage

Abstract:

Carbon injection projects introduce several operational, management and risk-assessment challenges, including leakage of carbon dioxide through fractures or derelict wells, reactivation of dormant faults, topographic subsidence/uplift, and contamination of existing groundwater resources. In this work we evaluate whether geomechanical measurements can be used within a stochastic estimation framework to characterize physical and geometric parameters of a system undergoing injection.

A numerical forward model built in COMSOL Multiphysics is used to solve the equations governing linear poroelasticity to compute the geomechanical signals (e.g. pressure, strain, tilt and displacement) produced during injection given a set of model parameters (e.g. Young’s modulus, Poisson’s ratio, and permeability). A hybrid Markov Chain Monte Carlo/multiobjective genetic algorithm is then used to iteratively generate a sequence of parameter estimates; distributed high performance computing is used to efficiently evaluate the computationally expensive forward model for each set of parameters. The set of posterior parameter estimates is then used to find the mean and uncertainty of each parameter subject to measurement limitations (noise, model error, spatial/temporal constraints).

We find that geomechanical measurements collected within the target formation can be used to accurately and efficiently estimate the physical parameters of the formation. We also observe that measurements taken in an overlying confining unit can be used to estimate the parameters of both the confining unit and target aquifer. This suggests that measurements made in the upper confining unit could mitigate drilling costs as well as reduce the risk of puncturing the confining unit. Using various combinations of synthetic measurements from the confining unit and target aquifer, we have also been able to resolve the geometry and physical parameters of heterogeneities analogous to fluvial gravel lenses and faults. Finally, we have found that when the model used in the inversion is conceptually flawed, systematic data fitting errors occur to alert practitioners to the problem. These numerical experiments demonstrate the potential of geomechanical measurements for characterizing coupled flow and deformation processes in a reservoir.