Régulation fluviale par commande prédictive avec modèle de connaissance embarqué

River regulation through predictive control with all on-board knowledge model

¹ Compagnie Nationale du Rhône
² Institut des Sciences et Techniques de l'ingénieur de Lyon

Abstract

All hydraulic structure designs now make use of mathematical models.

Knowledge based models use a discrete form of non-linear Barré-de-St-venant equations describing one dimensional free surface flows. Based on a structure of complex topological data, once calibrated, they permit simulating the hydraulic behaviour of rivers and calculating the levels and flows at each point.

The capacities of current computers now make it possible to use a knowledge-based model of hydraulic structures. Such a model embedded in a structure control computer permits implementing a robust predictive command capable of taking into account estimated future changes as trends, thus increasing control performance.

Different types of embedded model management can be used to formulate the control (scenario management, identification of an adaptative linear model, etc.).

The needs of overall management of volumes during floods are now leading to consideration on the possibilities of centralising the operation of a chain of developments. The feasibility of such a system has been studied for the dams of the Rhone downstream of Lyon.