Mathematical modeling of microalgal growth – hybrid and data-driven approaches for multi-scale prediction
Monday, 23 November, 2pm
Amphi Dorothy Hodgkin, ENS Paris-Saclay, Gif-sur-Yvette and online
Abstract: Microalgae have numerous applications, ranging from wastewater treatment to the production of proteins and biofuels. Developing these applications requires a better understanding and prediction of microalgal growth, particularly when moving from laboratory cultures to larger-scale systems.
This thesis combines experiments, mathematical modeling, and machine learning to address this challenge. Two multi-condition and multi-scale datasets were generated to study the growth of Chlamydomonas reinhardtii and Nannochloropsis oculata. Based on these data, hybrid modeling approaches combining ordinary differential equations and response surfaces, as well as sequential neural networks, were developed to predict microalgal growth under untested conditions. The ability of these methods to transfer predictions across scales was investigated using two laboratory culture scales as a proof of concept for scale-up. Finally, a theoretical analysis of microbial growth models provides further insight into the origin and limitations of some classical formulations.
Supervision
The thesis was prepared within the LMF, LuMIn, and later SATIE laboratories, under the supervision of Sakina Bensalem, Thomas Nowak, Matthias Függer, and Bruno Le Pioufle.
Jury
- Caroline Baroukh, INRAE Toulouse, Reviewer
- Francis Mairet, Ifremer Nantes, Reviewer
- Olivier Bernard, INRIA Sophia Antipolis, Examiner
- Hervé Duval, LGPM CentraleSupélec, Examiner