Date of Award
8-2019
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Environmental Engineering and Earth Sciences
Committee Member
Lawrence Murdoch, Committee Chair
Committee Member
Stephen Moysey
Committee Member
Scott DeWolf
Committee Member
Brian Dean
Abstract
Understanding the structure and material properties of hydrologic systems is important for a number of applications, including carbon dioxide injection for geological carbon storage or enhanced oil recovery, monitoring of hydraulic fracturing projects, mine dewatering, environmental remediation and managing geothermal reservoirs. These applications require a detailed knowledge of the geologic systems being impacted, in order to optimize their operation and safety. In order to evaluate, monitor and manage such hydrologic systems, a stochastic estimation framework was developed which is capable of characterizing the structure and physical parameters of the subsurface. This software framework uses a set of stochastic optimization algorithms to calibrate a heterogeneous subsurface flow model to available field data, and to construct an ensemble of models which represent the range of system states that would explain this data.
Many of these systems, such as oil reservoirs, are deep and hydraulically isolted from the shallow subsurface making near-surface fluid pressure measurements uninformative. Near-surface strainmeter, tiltmeter and extensometer signals were therefore evaluated in terms of their potential information content for calibrating poroelastic flow models. Such geomechanical signals are caused by mechanical deformation, and therefore travel through hydraulically impermeable rock much more quickly. A numerical geomechanics model was therefore developed using Geocentric, which couples subsurface flow and elastic deformation equations to simulate geomechanical signals (e.g. pressure, strain, tilt and displacement) given a set of model parameters. A high-performance cluster computer performs this computationally expensive simulation for each set of parameters, and compares the simulation results to measured data in order to evaluate the likelihood of each model. The set of data-model comparisons are then used to estimate each unknown parameter, as well as the uncertainty of each parameter estimate. This uncertainty can be inuenced by limitations in the measured dataset such as random noise, instrument drift, and the number and location of sensors, as well as by conceptual model errors and false underlying assumptions.
In this study we find that strain measurements taken from the shallow subsurface can be used to estimate the structure and material parameters of geologic layers much deeper in the subsurface. This can signicantly mitigate drilling and installation costs of monitoring wells, as well as reduce the risk of puncturing or fracturing a target reservoir. These parameter estimates were also used to develop an ensemble of calibrated hydromechanical models which can predict the range of system behavior and inform decision-making on the management of an aquifer or reservoir.
Recommended Citation
Hanna, Alexander Charles, "Stochastic Parameter Estimation of Poroelastic Processes Using Geomechanical Measurements" (2019). All Dissertations. 2478.
https://tigerprints.clemson.edu/all_dissertations/2478