Back
to main page of AQUANET
Project tasks
Task 1 - Project management and dissemination
This task aims at controlling project execution in order that the working flow proceeds according to schedule. Control of project
dissemination is also included in this task. Work flow control is done in a twofold way, including predictive and review control. For
predictive control management a detailed short/medium term planning of the work to perform in each task is made and reviewed in a
planning meeting, including the check that needed resources are or will be available when needed. For review control, recently
performed work within the project is reviewed in order to check whether results are according to planned. If not, adequate corrective
measures are decided for the period to come. A special attention will be devoted to deliver milestones at specified dates. In order to
ensure coordination among the different groups involved and implement this procedure, regular bi-monthly meetings with all the
groups involved in the project will be organised. Furthermore, there will be exchange of information using electronic means (email
and WWW) and more restricted work meetings where only sub-groups of the research team will be present.
The dissemination of the project results will done:
i) Through the publication of scientific papers;
ii) By creating and maintaining a WWW page;
iii) By organising a scientific seminars and a workshop. Contacts with majors irrigation water users in Portugal (EDIA,
FENAREG .- Federação Nacional dos Regantes de Portugal) will be maintained in order to disseminate project results.
Representatives of all partners will be involved in this task, in order to ensure an efficient communication with all the researchers
involved.
Task 2 - Canal models
Task description and Expected results
The objective of this task consists in making available to all researchers involved a mathematical model of the pilot canal that allows
testing in simulation of the different control and supervision algorithms to be developed. In this way, comparisons using a common
basis are possible. The model is based on the numerical integration of the Saint-Venant equations, with boundary conditions. It
should be flexible enough so that, not only the parameters may be changed but also canals with different structures may be
simulated. Experiments in the actual canal are needed in order to obtain data for parameter calibration. In this task, the study of the
best location for sensors based on observability properties of canal models will also be made. The expected result are thus a model of
the pilot canal implemented in MATLAB and the study of the impact of sensor location on model properties. The results of this task
will therefore be used in tasks 4, 5 and 6.
Task 3 - Canal refurbishment
The aim of this task is the refurbishment of the pilot canal, including:
-Refurbishment of the SCADA UPSs;
-Installation of the SCADA system in a convenient platform;
-Update of the SCADA presentation interface;
-Development of a common interface between the SCADA and C programs and MATLAB;
-Installation of flow sensors.
As a result, a software platform to interface control programs will be made available. This is an essential condition for the
experimental demonstration of the algorithms to develop in tasks 4, 5 and 6, and to the experimental demonstration in task 7.
Task 4 - Agent based control and decision
Inline with the overall goal of the project, this task will develop algorithms to support the co-operation between controllers within the
water distribution network. For that sake, the intelligent co-operative agent concept will be used attached to each controller at
specific network nodes. Thus, the development of work in this task will be made in two lines: (i) the co-operation mechanism
(negotiation, flow/volumes, timing, etc.) to achieve the global optimization under constraints; and (ii) the connection between the
local agent and the associated controller and the definition of the decisive “free” variables (e.g. level reference, maximum consumer
extraction, etc.).
An important tool is distributed model predictive control (MPC), where local decision centres (agents) exchange limited information in
order to coordinate themselves and achieve a near-optimal solution, but without the limitations and computational burden of
centralised multivariable MPC.
The expected results from this task are thus distributed decision and control structures based on local dynamic agents that allow
controlling a canal with limited access to global information, but that are able to meet performances, e. g. avoiding reflected waves
and canal instabilities.
The results will be evaluated both under a simulation environment (to be partial developed within the task) and through its
implementation in the experimental canal, in particular contributing to task 7.
Task 5 - Fault Detection and Isolation for Distributed Systems
This task refers to the development of a system capable of providing information related to the occurrence, place in the process and
size of faults. The research will be focused on the development of a robust FDI system based on robust observers, neural networks,
fuzzy systems, neuro-fuzzy strategies, knowledge based systems and fuzzy qualitative reasoning. A robust and reliable FDI scheme
will be able to distinguish the measurement process variables transient behaviours due to normal setpoint regulation from fault
effects, like sensor faults, actuator faults, process components faults and possibly controller faults. This will eventually facilitate the
reconfiguration required in an active fault-tolerant control system.
Within this task mainly model-based FDI techniques will be used. Therefore, models for the normal operation and for every possible
fault must be generated. Modelling of the canal system on operating conditions and faulty situations based on physical principles and
the definition of a bank of models together with a description of faulty situations by a logic programming environment. Some
analytical nonlinear models based on computational intelligence methods, as e.g. fuzzy, neural or neuro-fuzzy models, qualitative and
heuristic models will be developed for the experimental water delivery canal of the Núcleo de Hidráulica e Controlo de Canais from
University of Évora. After a fault has been detected in the process under supervision, the fault detection system will generate a signal
that will trigger a fault isolation approach responsible to locate and characterise the fault or faults being detected. The fault isolation
task will be performed using artificial neural networks. Furthermore, the FDI technique should deliver all the necessary information
towards controller reconfiguration online. Performance indexes will be defined and evaluated in order to assess the performance of
the FDI system during the on-line experiments, considering all the faults as abrupt, incipient and multiple simultaneous. The results
of this task will be used in task 6 (FTC) and task 7 (system integration).
Task 6 - Agent based fault tolerant control
Adopting an agent-oriented approach to fault-tolerant control means decomposing the problem into multiple-overlapping autonomous
parts (e.g. fault diagnosis and reconfiguration, or physical or functional partition of the process), that can act and interact in flexible
ways to achieve their set of objectives or to manage the dependencies that ensue from being situated in a common environment. The
overall set of multi-agents work together to deliver the overall functionality, in this case, a fault-tolerant system involving fault
diagnosis and control reconfiguration.
The key issue relates to the decomposition of the canal system (taken as a complex system). For this sake, methods based on
Bayesian Networks, Model Predictive Control (MPC) and Hybrid Systems will be considered. This results in an integrated flexible,
intelligent and scalable architecture for FTC with higher degree of reliability and performance.
The work focuses on the following topics:
1) Build both a Takagi-Sugeno (TS) fuzzy model and a hybrid model of the water
distribution system,
2) Apply and test algorithms for state estimation of both models,
3) Design and in-line implementation of a
Model Predictive Control (MPC) scheme with Fault Tolerant Control abilities based on the TS fuzzy and hybrid models. Alternatives
based on Bayesian Networks will also be considered. The aim is to design a MPC for fault-tolerant control (FTC) assuming that a
single model should represent each failure condition of the system. To avoid the exponential grow of the number of needed models
with the number of failure conditions, the state of the actual system will be approximated by a convex approximation of the local
models states. The use of fuzzy Takagi-Sugeno (TS) models and hybrid models constructed based on both expert knowledge and
input-output data will be adopted. Both quantitative and qualitative modeling will be used concurrently and, whenever feasible, the
models will be used for both FDI (in task 5) and control reconfiguration. As the experimental water delivery canal is a relatively slow
process, it allows the real-time implementation of a model predictive control. Based on the fault detection and isolation techniques
developed in the previous task, new controller configurations will be adopted, which will modify the controller in either structure or
parameters in order to safely continue operation under minimal quality and safety losses for any possible fault.
The following main results are expected from this task:
– Development of a systematic methodology to obtain the fuzzy and hybrid models of the water distribution system.
– Design and implementation of adequate model-based predictive control architecture for FTC using both fuzzy and hybrid models.
-Design and implementation of an agent-oriented Bayesian network for FTC purposes;
– Integration and testing of the FTC approaches both in simulation and at the experimental water delivery canal of the Núcleo de
Hidráulica e Controlo de Canais from University of Évora
Task 7 - Control integration and testing
The main objective of this task consists in the integration of the results from tasks 4, 5 and 6 in order to produce and test a software
package that embeds the principles of FTC and distributed multi-agent based control for controlling complex multipurpose water
delivery canal systems. The approach to follow in this task consists in selecting the most promising options that are complementary
to each other in order to reach the goal of the overall project objective: To develop decentralized controller networks for multipurpose
hydraulic open-channel systems where a network of local decision agents cooperate in order to achieve a sub-optimum solution with
fault tolerant capability.
As such, this task will receive results from 4, 5 and 6, in which alternative options have been considered and will concentrate in the
most promising ones that will work as building blocks of the controller architecture. The task also depends on task 3 that will provide
a software infrastructure where these blocks will be plugged, as well as the appropriate operational conditions of the pilot canal,
where the final system will be demonstrated
Task 8 - Final report
The aim of this task consists in the preparation of the final scientific report presenting an overall synthesis of the main achievements
of the project. It will be coordinated by representatives from all the partners. The main result is the comprehensive and critical
description of the agent based controller structure and its main design guidelines proposed for complex water delivery canal systems.