Phd Position : Unknown Input Observers For An Efficient Management Of Hydrographical Networks Based...

Universities and Institutes of France

France

June 2, 2022

Description

  • Organisation/Company: IMT - Institut Mines-Télécom
  • Research Field: Computer science › Database management Computer science › Modelling tools Computer science › Programming
  • Researcher Profile: First Stage Researcher (R1)
  • Application Deadline: 02/06/2022 21:00 - Europe/London
  • Location: France › douai
  • Type Of Contract: Temporary
  • Job Status: Other
  • 1 - Context and funding

    Hydrographical networks are large scale systems composed of natural rivers and artificial canals. They are used to answer Human's needs in terms of irrigation, navigation, drinking water, industry. As environmental systems, they are disturbed by several factors due to Human's activities and extreme climate events.

    In automation and artificial intelligence, control strategies have been designed to reject these disturbances and guarantee the management objectives (Lozenguez et al., 2018). It consists in controlling hydraulic devices (gates, weir, pumping stations), which allow to dispatch water to guarantee the needs and avoiding floods. For irrigation systems, it is useful to well estimate and predict withdraws; for navigation systems, to well estimate and anticipate flows due to rainfall (Sadowska et al. 2018; Gu et al., 2018; Hadid et al., 2020). Even if predictive control strategies have been designed (Segovia et al., 2021b)., it is still possible to improve their performances with a good estimation and prediction of uncontrolled factors. It is of the utmost importance especially in the current climate change context.

    Existing methods aims at estimating the unknown inputs/outputs based on H∞ observer (Bedjaoui et al., 2006), Kalman filter (Barthélémy et al., 2021; Bedjaoui et al., 2009; Bedjaoui et al., 2011), or Moving Horizon Estimation (MHE) technique (Guekam et al. 2021; Segovia et al., 2019). However, new advanced methods based on data-driven for the estimation and prediction of unknown inputs inputs (injections, withdrawal, leaks, infiltration and evaporation or water) are required (Langueh et al., 2020; Barbot et al., 2009). Based on better estimation, localization and prediction of the disturbances, the predictive control strategies could reach to an efficient management of hydrographical networks.

    Some benchmarks, already developed, will be used to achieve this research; a laboratory canal of 5m with 1 reach, canal with three 15m branches, 12m canal with flow from 0 to 150l/s, real irrigation and navigation systems located in the north and south of France.

    2 - Research project

    The design of observers (smart sensors) has to deal with the characteristics and complex dynamics of these systems, i.e. large scale, nonlinearity, variable delays, attenuation, etc. Given the inherently complex nature of hydrographical networks, designing observers by assuming linear dynamics may not be reasonable. Instead, the step aims first at designing unknown input observers by considering nonlinear dynamics, second at using this information to improve the predictive control of inland waterways. Based on the good estimation of unknown inputs, the model predictive control (MPC) framework will be employed to guarantee the management objectives, as it offers a number of features that are suitable to deal with hydrographical networks. The methods will be tested by using a simulator coupling MATLAB and SIC^2. The main objective consists in:

    i) proposing from real and simulation data, a realistic and usable model of a hydrographical network system (carry out a comparative study of the data- based approach (Moore K. J., 2019) and the physics-based approach);

    ii) achieving a follow-up of field tests on the proposed benchmarks to build a database on which new algorithms will be designed and tested;

    iii) developing some methods to tune and develop news modules in the Software SIC^2 to get digital twins of real systems,

    iv) designing a new approach of unknown input observers for the efficient control of hydrographical networks in centralized/decentralized/distributed approaches (Koenig et al., 2005),

    v) designing a new optimal model predictive control (Segovia et al., 2021a),

    vi) publishing contributions and results in conferences and scientific journals.

    Key-words: Large-scale networked systems, hydrographical networks, irrigation systems, inland waterways, Observer, Unknown input, Model Predictive Control.

    3 -Team supervision and PhD registration

    Location : CERI Systèmes Numériques, IMT Nord Europe (Douai) and IRSTEA Montpellier, France.

    Supervisors : Gilles Belaud (IRSTEA), Eric Duviella (IMT Nord-Europe)

    Selection process

    5 - How to apply

    Send by mail: CV, cover letter, M1 and M2 transcripts and recommendation letters (optional) to :

    - Gilles Belaud, IRSTEA (gilles.belaud@supagro.fr

    - Eric Duviella, IMT Nord-Europe (eric.duviella@imt-nord-europe.fr)

    Offer Requirements Skills/Qualifications

    4 - Candidate profile

    Required skills:

    - Background in programming (C, Python, Matlab)

    - Background in modelling, mathematical/data models, data-driven control systems, output analysis (observation) and optimization; Environmental systems, Water systems.

    - Good level of English (B2 level or higher).

    Contact Information
  • Organisation/Company: IMT - Institut Mines-Télécom
  • Organisation Type: Higher Education Institute
  • Website: https: // www. imt.fr/
  • E-Mail: eric.duviella@imt-nord-europe.fr
  • Country: France
  • City: Palaiseau
  • Postal Code: 91120
  • Street: 19 place marguerite Perey
  • Similar Jobs

    Add to favorites Read more...

    Phd Position: Abductive Reasoning With Minimal Sensing In A Home Environment

    leading home automation system vendors. consumption of technical systems The general objective of the thesis is to monitor human activities The domain of research of the thesis is knowledge representation and objective is to evaluate...

    Georgia Institute of Technology

    United States May 1, 2022

    Add to favorites Read more...

    AI/ML Engineer - ATAS - Open Rank

    Categories: Artificial Intelligence, Machine Learning The Georgia Tech Research Institute's Machine Learning and Analytics Branch is and capable of supporting, developing, and/or deploying machine learning and Additionally, an AI/ML Engineer performs research in Machine Learning...

    Add to favorites Read more...

    Robust Data-Driven Models And Algorithms For The Electric Vehicle Ecosystem

    Research Field: Computer science › Informatics learning models for predicting the energy consumption of electric vehicles in energy consumption models, methods to optimize the routing of vehicles to charging stations will be developed. algorithms at...

    Add to favorites Read more...

    2-Year Post-Doc

    Research Field: Engineering Title: Modeling of the operation of a plasma arc torch This post-doctoral work is part of a project with SAFRAN, a French This project, led by the Institute for Research on Ceramics...

    Add to favorites Read more...

    Data Modeling And Graph Neural Networks For Multimodal Film Analysis

    Location: France › Toulouse Domain: Multimedia Analysis, Machine Learning, Computer Vision, Data Science Université Toulouse 3 Paul Sabatier, Laboratoire IRIT i3s.univ-cotedazur.fr/~sassatelli/ 3. ### About Université Toulouse 3 and IRIT irit.fr/ , https: // www. Data...