Dynamic reduced-order models of integrated physics-specific systems for carbon sequestration

Abstract

$$\text {CO}_2$$ leakage from a deep storage reservoir into a shallow groundwater aquifer through a wellbore is simulated using physics-specific models conceptualized as nonisothermal two-phase flow in a deep storage reservoir and through the wellbore, and reactive transport in the aquifer. Supercritical $$\text {CO}_2$$ is injected into the reservoir subsystem with uncertain permeabilities of reservoir and injection strategies (rate and duration as decision variables). The simulated pressure and $$\text {CO}_2$$ saturation at an uncertain location are connected to the wellbore model as a boundary condition to output $$\text {CO}_2$$ flux to the reactive transport model in the shallow aquifer. Uncertainties are propagated from the deep reservoir model, to the wellbore leakage model, and eventually to the reactive transport model, thus contributing to system performance. To minimize the risk associated with the $$\text {CO}_2$$ leakage, we develop dynamic reduced order models for quantifying and managing the uncertainty propagation.