Missions :

Physics Informed Neural Networks (PINNs) for battery modeling using Doyle-Fuller-Newman (DFN) partial differential equations, alternatively known as the pseudo-two-dimensional (P2D) model. Using the flexibility and power of neural networks, PINNs offer a promising approach to solving complex PDEs like the DFN model for lithium-ion batteries. The advantages of this approach are: fast inference once trained, alternative method to solve complex PDE’s, usage on optimization tasks, real-time simulation, on-board/on-line models, etc.
Expected results:
Evaluation of tools, difficulties, approach/method and prepare for integration of complete DFN PDEs (P2D).
Proof of concept on part of the model: particle model.
Quantify differences in terms of accuracy and calculation time compared with our models already coded in Simulink and Comsol.
Preparation for integration of complete DFN PDEs (P2D).

Stage pour la validation d'un BAC+5

Compétences scientifiques : Machine learning

Connaissances : Génie Electrique, batteries

Moyens/Méthodes/Logiciels : Maitrise de Python et Matlab/Simulink

Commentaires libres : La maîtrise de Python et Matlab/Simulink est indispensable. Des connaissances en machine learning sont nécessaires. Les connaissances générales en génie électrique et batteries sont un plus.