Post-doctoral position in design and evaluation of innovative control strategies for green hydrogen production
The production of green hydrogen through electrolysis is one of the relevant solutions for decarbonization and energy sustainability. However, it faces major challenges related to the variability of renewable energy sources such as wind and solar. This variability can accelerate the degradation of electrolyzers, increasing production costs and reducing their lifespan. To ensure competitive production of decarbonized hydrogen, it is essential to find a balance between the optimal use of decarbonized electricity and the maximization of electrolyzer operating time.
Recent studies, such as those by Y. Li et al. [1] and J. Wang et al. [2], have explored control strategies to optimize hydrogen production from renewable energies. Additionally, approaches based on artificial intelligence, such as those by J. Li et al. [3] and Yadollahi et al. [4], have shown their effectiveness in managing energy systems.
Within a team composed of about twenty experts (hydrogen systems modeling, operations research, machine learning, programming), this post-doctoral project will focus on demonstrating the feasibility and relevance of intelligent control strategies to reduce system degradation while optimizing green hydrogen production.
Your missions will be:
- To conduct a literature research and interviews with other CEA research in order to refine the definition of the envisioned case study with relevant parameters and data.
- To create a simulation model for the case study, which should be based on existing component models [5], and should include relevant reference control strategies to serve as a benchmark.
- To design and implement new control strategies based on machine learning and reinforcement learning algorithms to optimize the control of electrolyzers. These control strategies will be designed to reduce electrolyzer degradation, maximize energy efficiency and minimize production costs.
- To perform an in-depth analysis and comparison of possible control strategies on the case study, especially focusing on their performance regarding electrolyzer degradation.
- To publish results in scientific journals and contribute to cooperation with French and International academics. Publication of some project results as Open Data and/or Open Source will also be possible.
The work will be conducted in a collaboration between teams located at the National Institute for Solar Energy (Le Bourget du Lac, Savoie), and at CEA LITEN in Grenoble. You will fully benefit from the dynamic working environment at CEA, including possibilities for scientific collaboration at the national and international level. The position is located in Le Bourget du Lac, with possibilities for partial remote work.
Your profile:
You have a PhD in applied mathematics or control, with skills in scientific programming and a strong interest in energy.
You have experience of the following domains:
- Design of multi-level control strategies for complex systems (PID, rule-based, MPC) - required
- Understanding of energy system modeling and simulation – recommended
- Understanding of physics-based modeling of electrolyzer systems, especially considering degradation phenomenon - recommended
- Machine Learning, Deep Learning and/or Reinforcement Learning - recommended
You are proficient with at least one of Python / Modelica / Simulink / FMU standards, and are willing to learn the others. Being able to work in various computing environments (Windows, Linux).
You are fluent in English. You bring creativity, enjoy teamwork and can undertake research projects autonomously.
How to apply ?
In addition to providing a resume and motivation letter (max. 1 page - directly related to the offer), you are encouraged to reach out directly to Dr. Mathieu Vallée and Dr. Mohammed-Ali Jallal (firstname.lastname@cea.fr).
References
[1] Y. Li et al., “Exploration of the configuration and operation rule of the multi-electrolyzers hybrid system of large-scale alkaline water hydrogen production system,” Applied Energy, 2023, doi: 10.1016/j.apenergy.2022.120413.
[2] J. Wang, L. Kang, and Y. Liu, “Optimal design of a renewable hydrogen production system by coordinating multiple PV arrays and multiple electrolyzers,” Renew. Energy, vol. 225, no. January, p. 120304, 2024, doi: 10.1016/j.renene.2024.120304.
[3] J. Li, Z. Jiang, Z. Chen, J. Liu, and L. Cheng, “CuEMS: Deep reinforcement learning for community control of energy management systems in microgrids,” Energy Build., vol. 304, no. November 2023, p. 113865, 2024, doi: 10.1016/j.enbuild.2023.113865.
[4] Z. Yadollahi, R. Gharibi, R. Dashti, and A. Torabi Jahromi, “Optimal energy management of energy hub: A reinforcement learning approach,” Sustain. Cities Soc., vol. 102, no. January, p. 105179, 2024, doi: 10.1016/j.scs.2024.105179.
[5] F. Nepveu, S. Mathonniere, G. Enee, N. Lamaison. Presentation, validation and application of the Energy Process library. The 15th International Modelica Conference, Oct 2023, Aix-La-Chapelle, Germany. hal:cea-04214202f
