Intelligent control and estimation for floating wind turbines

  • Posted on: 6 May 2021
  • By: salah.laghrouch...
  • Updated on: 7 May 2021
Type recrutement: 
Sujet de thèse
Section: 
61 (Génie informatique, automatique et traitement du signal)
63 (Génie électrique, électronique, photonique et systèmes)
Entité et lieu: 
FEMTO-ST, UTBM
Urgent ?: 
oui
Détails: 

Host laboratory: FEMTO-ST Research Institute - Belfort

Thesis director (contact) : Dr. salah LAGHROUCHE – salah.laghrouche@utbm.fr

Co-advisors: Dr. Daniel DEPERNET

Funding: National Agency for Research (ANR)

Doctoral School: SPIM (Engineering Sciences and Microtechnologies), speciality Automation

Keywords: floating wind turbines, robust and adaptive control, deep learning, learning-based estimation, marine renewable energy

Scientific context:

One of the main challenges of floating wind turbines is the development of specific robust and adaptive control systems [McK_16]. Indeed, floating wind turbine systems are multivariable underactuated systems where the number of degrees of freedom (DOF) exceeds the number of control inputs (blade pitch, generator torque). Thus, the limited number of actuators makes it difficult to minimize structure motion while ensuring maximum power extraction without issuing conflicting commands to actuators or affecting other DOFs due to couplings. This particularity of the floating wind turbine implies to bring nontrivial answers to obtain the best compromise that guarantees the stability of the wind turbine and the maximization of the extracted power. In this context, very few viable control solutions exist, i.e., with an optimal production making the wind turbine economically profitable and structurally sustainable. It should be noted that the problem of control of floating wind turbines has been addressed by the automation community only for a few years and mainly by linear approaches, the use of nonlinear control approaches under realistic conditions, with consideration of the various couplings of the system, is very recent and not yet completed. Indeed, if we consider a model that represents the floating wind turbine with fidelity, it will necessarily be nonlinear and uncertain. In this case, the solution is much more complex. Hence the need to synthesize intelligent, robust and optimal control techniques for these systems in order to achieve good levels of performance in all operating areas.  

The PhD thesis is part of the ANR CREATIF (Control and REAl TIme simulation of Floating wind turbines and integration to grid) project which aims to implement a new "real-time" simulation tool of the "Hardware-In-The-Loop" type offering complete interaction models between the various components of a floating wind turbine, in order to: 1) develop new control and estimation architectures based on non-linear and deep learning approaches, which are efficient over a large operational domain for both energy production and wind turbine stabilization, and relatively simple to tune; 2) optimize the architectures and sizing of the energy conversion chains and their integration into the grid on technical-economic criteria. The project partners cover all the components and systems involved in the floating wind turbine process: wind turbine dynamics under the combined action of swell and wind; energy conversion chain and grid; control strategies; hardware simulation and Power-hardware-in-the-Loop.

Description of the thesis:

The first objective of the thesis is to synthesize novel robust, adaptive control and estimation solutions valid over the three operating zones of the floating wind turbine with a particular interest in the 1.5 and 2.5 transition zones. We plan to use adaptive control based on deep learning [Beno14, Shi19, Fuen20]. The effectiveness of this approach lies in the fact that these controllers use a basic, albeit partially known, physical model of the system and add a learning layer to compensate for the unknown part of the model. Through this combination, one could take advantage of the model-based design with its stability characteristics and add the advantages of the free learning model with its fast convergence and robustness to uncertainties. 

Wind field reconstruction from LiDAR (Light Detection And Ranging) data has been extensively addressed in the context of fixed wind turbines [Towe14, Schli16]. However, very little work exists for floating wind turbines. To our knowledge, the only work that exists is that of EDF Energy R&D UK Center. To obtain a valid measurement on a moving support (float), stabilization gyroscopes are used to limit the movements of the LIDAR. The motions and displacements of the array are captured with accelerometers to compensate the effects on the wind measurement using post-processing. The second objective of the thesis is to develop learning-based state estimators to predict the effects of wind. The novelty of this work lies in the fact that the LIDAR motions will be reconstructed and compensated based on the wave field prediction (work done by the LHEEA, ECN) and an adequate modeling of the floating structure response. This will make it possible to get rid of the use of gyroscopes and accelerometers and thus reduce the costs while increasing the reliability of the device.

The third objective of the thesis will be to test and validate experimentally the control, estimation and optimization techniques on two experimental platforms: 

- A "Power hardware in the loop" test bench at FEMTO-ST, Belfort. This system will allow the integrated and real time simulation of the complete system with the dynamics of the floating wind turbine numerically simulated and the energy conversion chain part using real power devices. 

- A device evolving in a wave tank located at the LHEEA (ECN Nantes) emulating the operation of a floating wind turbine. 

Presentation of the host laboratory:

The FEMTO-ST Institute (Franche-Comté Electronique Mécanique Thermique et Optique - Sciences et Technologies, UMR 6174) is a joint research unit under the quadruple supervision of the University of Franche-Comté (UFC), the National Centre for Scientific Research (CNRS), the École Nationale Supérieure de Mécanique et des Microtechniques (ENSMM) and the University of Technology of Belfort-Montbéliard (UTBM). Today, the FEMTO-ST Institute has 7 scientific departments and more than 700 members. The doctoral thesis will take place within the SHARPAC team of the Energy Department of the FEMTO-ST Institute in the premises the UTBM university site in Belfort. It will start in September 2021 for a period of 3 years.

Thesis funding:

This thesis will be funded by the National Agency for Research (ANR). The doctoral student will be registered at the UBFC (University of Bourgogne Franche-Comté) and at the SPIM doctoral school (Engineering Sciences and Microtechnologies) in the speciality Automation.

Candidate profile:

The candidate will have to demonstrate a strong motivation for scientific research and very good level of English language skills. He or she will have to demonstrate a great rigour in work, method, autonomy, and ease in experimenting, analysing and presenting data. He or she will have to hold a Master 2 or equivalent degree with very good academic results and to know the main notions of electrical engineering, control and industrial computing.

List of documents to be provided:

  • CV
  • A letter of motivation
  • Academic transcript and ranking of Master 1 and 2
  • Recommendation letter(s)

The application should be sent by e-mail to: salah.laghrouche@utbm.fr

The closing date for application is: 30/05/2021