Bridging model-based control and machine learning for the prognostic and health management of wind...

Updated: 3 months ago
Job Type: FullTime
Deadline: 15 May 2021


The energy sector is undergoing a profound transition to meet climate objectives while ensuring global access. On the energy production side, renewable energy sources such as wind and solar are gaining in popularity. Out of all the renewable energy alternatives, wind energy is the most developed technology worldwide [1].

In order to ensure reliability, safety and efficiency of wind turbines, thus avoiding expensive unplanned maintenance and down time, fault prognostic as well as post-prognosis decision-making for health management (PHM) is fundamental [2]. As a consequence, the design of PHM frameworks has gathered more and more attention and significant research effort is being taken separately by two distinct research communities: the model-based control community and the machine learning community.

In this context, model-based control methods have been extensively developed to predict faults by considering a physical or a mathematical description of the system’s behavior. A major advantage of these methods consists on their versatility and their ability to provide robustness against uncertainty and disturbance, safety guarantees and cost-oriented control. However, in practice, it is very difficult to accurately model all the possible wind conditions affecting wind turbines and to develop control strategies with safe and robust performance. Here, a major challenge is to provide implementable solutions that require low online computational and memory requirements.

Due to the increasing availability of data as well as new computation, sensing and communications capabilities, the last years have witnessed an enormous interest in the use of machine learning techniques for PHM of wind turbines.

Thereby, each approach has its own advantages and drawbacks and each community has developed its own concepts, tools and techniques, guided by their different modeling paradigms and backgrounds. Nowadays, it seems that the development of a comprehensive PHM solution requires understanding and synergistically integrating these control and machine learning techniques in an efficient and optimal manner [3, 4]. This is a promising research track that will be further catalyzed by the emerging groups such as "Automatique et IA" group, supported by GDR MACS and the recent IFAC workshop “Machine Learning meets Model-based Control”.


The objective of this thesis is to integrate machine learning techniques with model-based control theory for the prognostic and health management of wind turbines in presence of unpredictable wind conditions. A first part will be dedicated to the study of unifying frameworks and generic methods allowing the integration of machine learning approaches within the framework of the control community. This study should take into account the fundamental challenges related to control properties such as stability, convergence, constraint satisfaction and performance under uncertainty, raised by this combination. A second part will focus on developing a comprehensive PHM solution for wind turbines, that exploits the important synergies between model-based control and machine learning, and that contributes to solve the identified challenges specific to this integration and to wind turbines in general.

- First year:

  • Bibliographic study on model-based control, machine learning methods and existing frameworks for their integration.
  • Study of wind turbines and their existing PHM solutions.
  • Study of the modeling and data requirements of wind turbines
  • Preparation of wind turbines data.

- Second year:

  • Development of fault prognosis solution based on available data.
  • Improvement of the fault prognosis solution using model-based control tools.
  • Investigation of feasible post-prognostic actions and estimation of their costs (maintenance schedule, reconfiguration, control actions, etc.).
  • Development of post-prognostic decision-making tools.

- Third year:

  • Application of the proposed methodology and validation of the framework.
  • Thesis manuscript and valorization.

KEYWORDS: model-based control, machine learning, prognostic and health management, wind turbines.


[1] Simani, S., & Farsoni, S. (2018). Fault Diagnosis and Sustainable Control of Wind Turbines: Robust data-driven and model-based strategies. Butterworth-Heinemann.

[2] Kusiak, A., & Li, W. (2011). The prediction and diagnosis of wind turbine faults. Renewable energy, 36(1), 16-23.

[3] Tidriri, K., Chatti, N., Verron, S., & Tiplica, T. (2016). Bridging data-driven and model-based approaches for process fault diagnosis and health monitoring: A review of researches and future challenges. Annual Reviews in Control, 42, 63-81.

[4] Tidriri, K., Verron, S., Tiplica, T., & Chatti, N. (2019). A decision fusion based methodology for fault Prognostic and Health Management of complex systems. Applied Soft Computing, 83, 105622.

SCHOLARSHIP: 3 years scholarship awarded by the Doctoral School


Start date: October 2021.

Application deadline: 15 May 2021.

Research team: SAFE

Send CV, transcripts from the last two years of study and cover letter to: and

View or Apply

Similar Positions