%0 Generic
%A Thai, Tran Hong
%D 2005
%F heidok:5662
%R 10.11588/heidok.00005662
%T Numerical Methods for Parameter Estimation and Optimal Control of the Red River Network
%U https://archiv.ub.uni-heidelberg.de/volltextserver/5662/
%X In this thesis efficient numerical methods for the simulation, the parameter estimation, and the optimal control of the Red River system are presented. The model of the Red River system is based on the Saint-Venant equation system, which consists of two nonlinear first-order hyperbolic Partial Differential Equations (PDE) in space and in time. In general a  system of equations  of  this type can not be solved analytically. Therefore I choose  a numerical approach, namely the Method Of Lines (MOL) combined with the Backward  Differentiation Formulae (BDF) method implemented in the solver DAESOL, which is developed at IWR (University of Heidelberg), for the solution of the Saint-Venant equation for the simulation of the Red River system.   In a river,  there are  geometrical and hydraulic parameters, e.g. friction coefficients, river bed slope, etc., which are very expensive or -- even worse -- impossible to measure.  The Red River system is a large system, therefore  the number of the unknown parameters is also large, thus manual parameter estimation, as often done by hydrologists, does not work effectively. To overcome this problem and to come up with accurate parameter values, we estimate these parameters by solving a corresponding least-squares problem. This high dimensional nonlinear constrained optimization problem is solved by applying a special reduced Gauss-Newton method implemented in the software packages for parameter estimation PARFIT and FIXFIT. Based on the code  FIXFIT we have developed a numerical tool for solution of simulation and parameter estimation problems for river flows that are modeled  by hyperbolic  PDE. Using the validated model, we  formulate an optimal control problem  for preventing floods in the Red River lowland. This problem is solved by the direct multiple shooting method within the software package MUSCOD-II. We  propose an online optimization approach with a Nonlinear Model Predictive Control (NMPC) technique to reduce the maximum flood level at Ha Noi by controlling the  water discharge at the output of the reservoir Hoa Binh. The potential of our approach is demonstrated  in real-life test cases corresponding to the flood season in the year 2000.