PNNL

Abstract

A series of numerical test cases reflecting broad and realistic ranges of geological formation properties was developed to systematically evaluate and compare the impacts of those properties on geomechanical responses to CO2 injection. A coupled hydro-geomechanical subsurface transport simulator, STOMP (Subsurface Transport over Multiple Phases), was adopted to simulate the CO2 migration process and geomechanical behaviors of the surrounding geological formations. A quasi-Monte Carlo sampling method was applied to efficiently sample a high-dimensional parameter space consisting of injection rate and 14 subsurface formation properties, including porosity, permeability, entry pressure, irreducible gas and aqueous saturation, Young’s modulus, and Poisson’s ratio for both reservoir and caprock. Generalized cross-validation and analysis of variance methods were used to quantitatively measure the significance of the 15 input parameters. Reservoir porosity, permeability, and injection rate were found to be among the most significant factors affecting the geomechanical responses to the CO2 injection. We used a quadrature generalized linear model to build a reduced-order model that can estimate the geomechanical response instantly instead of running computationally expensive numerical simulations. The injection pressure and ground surface displacement are often monitored for injection well safety, and are believed can partially reflect the risk of fault reactivation and seismicity. Based on the reduced order model and response surface, the input parameters can be screened for control the risk of induced seismicity. The uncertainty of the subsurface structure properties cause the numerical simulation based on a single or a few samples does not accurately estimate the geomechanical response in the actual injection site. Probability of risk can be used to evaluate and predict the risk of injection when there are great uncertainty in the subsurface properties and operation conditions.