Application des algorithmes genetiques pour le calibrage d'un modele hydrologique couple (French text).

摘要

A conceptual surface runoff global model, built with two reservoirs and coupled with the Green and Ampt infiltration model (1911) was developed. This rainfall-infiltration-runoff model (PIR) expresses the physical behavior of the infiltration phenomena. The PIR model has three calibration parameters, two of which are related to the Green and Ampt model (those being hydraulic conductivity K and B) and the last one is the threshold height of surface runoff (h_sm). Experiments were done in the laboratory to collect some rainfall-runoff measurements. The K and B infiltration parameters were estimated using two realized methods of experimental measurements fitting. The first one calculates the infiltration capacity by numerical time derivation of the infiltration heights. The second method calculates the infiltration capacity by the analytical derivation of the fitted cumulated infiltration height function. Since the latter performed the best results, those ones were retained for the validation of the parameters calibration, in this work.; The PIR model calibration was realised, from the experimental data, by the simplex method. This method had diverged because of the complex shape of the objective function characterized by plates and valleys. The genetic algorithm was used to calibrate the PIR model from experimental and synthesized data. This method had successfully solved the optimization difficulties. K and B parameters obtained by model calibration, from experimental measurements, were compared to the parameters obtained from experimental data fitting. The K and B PIR model parameters, which were obtained from synthetic calibration were compared to the preset values of parameters. In fact, the optimization results, achieved in this study, showed that the genetic algorithm was efficient and robust in our case.