Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → Espagne → 1990 → Modelos hidrológicos de ayuda a la decisión en tiempo real basados en técnicas de inteligencia artificial

Universidad Politécnica de Madrid (UPM) 1990

Modelos hidrológicos de ayuda a la decisión en tiempo real basados en técnicas de inteligencia artificial

Garrote de Marcos, Luis

Titre : Modelos hidrológicos de ayuda a la decisión en tiempo real basados en técnicas de inteligencia artificial

Auteur : Garrote de Marcos, Luis

Université de soutenance : Universidad Politécnica de Madrid (UPM) E.T.S.I. Caminos, Canales y Puertos

Grade : Tesis (Doctoral) 1990

Résumé
The thesis presents two knowledge representation environments : SIM and CYRAH, which are to be used for real-time decision support in flood forecasting in Mediterranean basins. The decision making during a flood event involves a number of steps that clearly exceed the possibilities which are currently offered by classical mathematical hydrologic models. The decision usually takes place in a professional environment where uncertainty sets practical limitations to the use of simulation models. Knowledge based systems offer the possibility of representing the actual reasoning and judgements made by the experts to build a model that embodies the professional knowledge usually applied in these cases. First, a general approach for decisión support systems during floods is presented. Flood forecasting in the Mediterranean área is greatly influenced by the meteorological, topographical and geomorphological features of its watersheds. Short rivers, steep slopes, deforestation and torrential rains are the main causes for a fast and intense response to rainfall in the zone and the consequent flood. A warning system must meet strong requirements : prompt response of relative accuracy under demanding conditions, such as unpredictable meteorological situations or ignorance of the basins’ physical behaviour. An expert system approach is proposed as an aid in decisión making, due to the need for handling qualitative knowledge and criteria about physical phenomena. A methodology for building this kind of knowledge-based systems is presented. Physical processes should be properly represented, and thus the knowledge base and inference engine of this type of expert system must be specially adapted to meet this requirement. Therefore, the proposed architecture for these knowledge based systems integrates the professional reasoning paradigms used by hydrologists, such as traditional simulation models, together with symbolic problem solving paradigms such as rules and frames. The decisión support system is conceived as an inteUigent environment capable of automatically identifying the problems that might occur consistently with the set of data received during a flood event. The ability to direct the search for problems with a rough engineering model and then refine and adjust the evalualion using numerical analysis is a classical approach to problem solving in engineering. This approach is the basis for the overall structure of the environments within the scope of real time flood forecasting. Much emphasis is placed on the fact that the inputs and the physical behaviour of the system are only partially known and therefore the uncertainty on the data sould be assumed. Two original models are presented as an outcome of the research : SIM and CYRAH. Both are intermediate applications that stand between the classical numeric simulation systems, based on mathematical models, and the artificial intelligence systems, based on symbolic knowledge representation and inference procedures. In order to represent the behaviour of the physical system a reasoning line has to be assumed, but most of the knowledge handled by the environments is intended to be declarative ; that is, defíned by the user to adapt to his way of understanding the decisión making process. The SIM environment (Simulación e Inferencia con Modelos) is capable of making knowledge-based simulations of the behaviour of the basins. It represents the reasoning about the behaviour of the physical system using two numeric simulation models : SAVEL and REBOLSA, integrated in a knowledge representation environment : SAR. SIM is a knowledge-based environment that lets the user define the criteria for handling ill-calibrated hydrological models under incomplete Information. The CYRAH environment (Cálculo Y RAzonamiento Hidrológico) integrales the simulation module SIM in a basic reasoning line about the flooding process : In a first phase, the current situation is analized. The expert is able to focus his attention only in the cases that are or might be problematic in the futiu"e. His knowledge of past events enables him to make conjectures on plausible evolutions of the meteorológica ! situation and evalúate roughly the response in the different basins to future rains. As a result of his analysis, a limited number of cases are selected for further attention. In a second phase, the SIM environment is used lo estímate discharges and levéis in those cases. The uncertainty in the data and the descriptors of the physical structure is overeóme by repeating the simulations using representative combinations of inputs and inlerpreting the different outpuls obtained. The chances of problems arising in the near future are deduced in a third phase from the discharges and levéis estimaied.

Mots Clés : Robotics and Industrial Computing Mathematics Geography

Présentation

Version intégrale (13 Mb)

Page publiée le 6 février 2015, mise à jour le 14 mars 2017