Solving the Estimation-Identification Problem in Two-Phase Flow Modeling
Stefan Finsterle and Karsten Pruess
Water Resources Research 31(4), 913-924, April 1995
Lawrence Berkeley National Laboratory, Earth Sciences Division
University of California, Berkeley, CA 94720
Abstract. A procedure is presented to solve the estimation-identification problem in two-phase flow modeling. Given discrete observations made on the system response, an optimum parameter set is derived for an appropriate conceptual model by solving the inverse problem using standard optimization techniques. Subsequently, a detailed error analysis is performed, and nonlinearity effects are considered. We discuss the iterative process of model identification and parameter estimation for a ventilation test performed at the Grimsel Rock Laboratory, Switzerland. A numerical model of the ventilation drift and the surrounding crystalline rock matrix is developed. Evaporation of moisture at the drift surface and the propagation of the unsaturated zone into the formation is simulated. A sensitivity analysis is performed to identify the parameters to be estimated. Absolute permeability and two parameters of van Genuchten’s characteristic curves are subsequently determined based on measurements of negative water potentials, evaporation rates, and gas pressure data. The performance of the minimization algorithm and the system behavior for the optimum parameter set are discussed. The study shows that a field experiment conducted under two- phase flow conditions can be successfully reproduced by taking into account a variety of physical processes, and that it is possible to reliably determine the two-phase hydraulic properties that are related to the given conceptual model.