Back to main page of AQUANET

IDEA Project Summary

The scarcity of fresh water is becoming one of the most important environment constraints with a major impact on economic development and the quality of life in the South of Europe, including Portugal. The AQUANET project aims at contributing to mitigate these problems by the development of advanced control methodologies for optimizing the management of water conveyance and delivery in multipurpose open-channel systems, with the goal of minimizing the use of energy and water spills. Besides this major socioeconomic motivation, contributing to the solution of this problem implies the need to push forward research in several aspects of control, as explained hereafter. In short, the overall goal of the project is the development of decentralized controller networks for multipurpose hydraulic open-channel systems where a network of local decision agents cooperate in order to achieve a near optimum solution with fault tolerant capability.

The systems are large, spatially distributed dynamic plants. Usually, they form branched networks where the basis for modeling each branch are the Saint-Venant equations, a set of nonlinear first order partial differential equations of hyperbolic type that embody conservation of mass and of moment. Together with suitable boundary conditions, their solution models the water level along the different branches. The control system to design receives data in real time from the sensors (water levels at selected points) and decides, at each sampling time interval what should be the value of the command for the actuators (gate levels).
The major problems to confront are:
1) The distributed character and large scale of the system.
2) The uncertainty in the knowledge of the dynamics and its complexity.
3) The existence of operational constraints.
4) The existence of scheduled unexpected disturbances induced by water consumption at the turnouts.
5) The inherent delay between control action and result.

Similar problems have been, and currently are, in consideration among the World scientific community. Besides the fact that the problem is by no means completely solved (thereby requiring new contributions), what distinguishes the approach proposed in this project is the combination of the following features:

1) The use of distributed agents that cooperate in order to achieve a near optimum control solution that embeds fault tolerance;
2) The modularity of the basic algorithms, allowing their interconnection in order to produce solutions that are scalable and reconfigurable with respect to the type of hydraulic canal system considered.

The above approach will be pursued based on the following methods. Methods for decentralized control allow to coordinate networks of local control agents. An approach based on Model Predictive Control (MPC) with coordination achieved through augmented Lagrangian will be followed. MPC has the advantage of providing a flexible control optimization tool whose basic version may be modified in order to incorporate constraints and fault tolerant features, as well as reconfiguration and adaptation. In order to achieve reconfiguration and adaptation, a natural way is to explore methods based on switched multiple models.
Due to the distributed nature of the plant an important issue is sensors location. This aspect will be addressed by studying the resulting controllability and observability of the lumped parameter models used for control design. A distinctive feature of the project is the availability of a large scale pilot plant, where real tests will be conducted. This is the experimental canal of the Núcleo de Hidráulica e Controlo de Canais from University of Évora with a total length of 150 m and designed for 90 l/s. At this scale, this is one of the three unique facilities of its kind worldwide.

A short description of the pilot canal may be found here pdf


Achieving the global objective of this project requires an interdisciplinary team that is actually gathered in this project. The research team has competencies on the core issue of Control (INESC-ID, IDMEC/IST), operational control and design of open canal systems (University of Évora), intelligent agent-based decision algorithms (UNINOVA, IDMEC/IST), software design for control (University of Évora, INESC-ID and UNINOVA), fault detection and isolation as well as fault tolerant control techniques using neuro-fuzzy, hybrid, and multi-agent systems (IDMEC/IST), dynamic modeling and simulation of industrial and canal systems (INESC-ID and UNINOVA). The main scientific contributions from this project are:

1) New design procedures for water delivery canal network control, combining distributed agents and fault tolerance.

2) New algorithms for model predictive control to embody a combination of distributed agents and fault tolerance.

3) New algorithms of adaptation for model predictive control based on multiple models, for the class of problems considered.

4) New fault tolerant control algorithms based on hybrid systems modeling.

5) New optimization-coordination control algorithms for distributed systems.