Decentralized adaptive control for a water distribution system

This paper considers the design of a decentralized adaptive controller for a water distribution system. This system consists of a 40-kilometer long canal connected to three secondary canals and four main reservoirs supplied by a 7 m/sup 3/ s/sup -1/ river (the Bourne river). More than 20 pumping stations are distributed along the canals, bringing water to the agricultural plain of Valence (South-East of France). The system is currently manually operated through flow control gates. The author's medium-term objective is to automatically operate this system and to improve water distribution efficiency and safety. In terms of control the primary goal is to satisfy the water demand at each pumping station, while guaranteeing a minimum water level, overflow avoidance and minimum wastes of water, in each section and reservoir of the system. The author's approach is divided into two main stages. An on-line identification procedure based both on a multi-input multi-output parametrized model and Kalman filtering. A state-space predictive control scheme using a Kalman state estimator defined on the basis of the on-line identified model. The author shows how this control scheme can be decentralized in case of large-scale water distribution systems, such as the author's process.<>