Trainee Project
Title:
Domain adaptation for lifetime prediction of industrial batteries
Dates:
2024/04/02 - 2024/09/30
Supervisor(s):
Other supervisor(s):
CLAUSEL Marianne (marianne.clausel@univ-lorraine.fr)
, BERTIN Clément
Description:
Worldwide demand for electrochemical batteries is growing all the time. This is mainly due to the emergence of hybrid and electric vehicles (Hybrid-
Electric Vehicle, Plug-in Hybrid Electric Vehicle and Battery-Electric Vehicle) on the one hand, and the energy storage market linked to renewable
energies and power grid management on the other.
SAFT is particularly active in this context, as a pioneer in the deployment of lithium batteries. SAFT produces lithium-ion batteries in Poitiers, Nersac
and Bordeaux. This internship, funded by SAFT, will take place at CRAN in Vandoeuvre-lès-Nancy.
Description: One of the limiting aspects of batteries is that they age over time. Depending on their composition and conditions of use, their performance
degrades until they are considered unfit for their intended purpose. This time corresponds to their service life. For manufacturers, knowledge of the state
of ageing is a major issue, and a key factor in determining the price of a battery. It also helps to understand the causes of ageing, and possibly to improve
cell design. To quantify the service life of a new type of battery, manufacturers carry out a series of ageing tests under controlled experimental conditions
on a batch of batteries. Degradation is quantified using state-of-health (SoH) indicators, typically capacity or internal resistance. These tests are costly
and time-consuming, so only a few batteries are tested. These data are then used to model the typical time evolution of SoH.
The aim of this internship is to capitalize on a thesis carried out at Saft, which studied the life degradation trajectories of Li-ion batteries using Gaussian
processes. Building on this work, the student will model the life degradation of a battery that has not yet been studied, focusing on the transfer of
hyperparameters from existing models.
Study plan: Implementation of thesis work ([1], [2]) on the modeling of Li-ion battery life degradation based on Gaussian processes on a new battery
dataset. Exploration of context-sensitive domain adaptation methods.
Development of a life degradation model for a new type of battery based on this preliminary study.
Implementation of an evaluation process for the model thus developed.
References :
[1] B. Larvaron, M. Clausel, A. Bertoncello, S. Benjamin, G. Oppenheim, Chained Gaussian processes to estimate battery health degradation with
uncertainties. Journal of Energy Storage, 67, 107443, 2023
[2] B. Larvaron, M. Clausel, A. Bertoncello, S. Benjamin, G. Oppenheim, Chained Gaussian processes with derivative information to forecast battery
health degradation. Journal of Energy Storage, 65, 107180, 2023
[3] Zhou, Zihao, et al. "Few-Shot Cross Domain Battery Capacity Estimation." Adjunct Proceedings of the 2021 ACM International Joint Conference on
Pervasive and Ubiquitous Computing and Proceedings of the 2021 ACM International Symposium on Wearable Computers. 2021.
Electric Vehicle, Plug-in Hybrid Electric Vehicle and Battery-Electric Vehicle) on the one hand, and the energy storage market linked to renewable
energies and power grid management on the other.
SAFT is particularly active in this context, as a pioneer in the deployment of lithium batteries. SAFT produces lithium-ion batteries in Poitiers, Nersac
and Bordeaux. This internship, funded by SAFT, will take place at CRAN in Vandoeuvre-lès-Nancy.
Description: One of the limiting aspects of batteries is that they age over time. Depending on their composition and conditions of use, their performance
degrades until they are considered unfit for their intended purpose. This time corresponds to their service life. For manufacturers, knowledge of the state
of ageing is a major issue, and a key factor in determining the price of a battery. It also helps to understand the causes of ageing, and possibly to improve
cell design. To quantify the service life of a new type of battery, manufacturers carry out a series of ageing tests under controlled experimental conditions
on a batch of batteries. Degradation is quantified using state-of-health (SoH) indicators, typically capacity or internal resistance. These tests are costly
and time-consuming, so only a few batteries are tested. These data are then used to model the typical time evolution of SoH.
The aim of this internship is to capitalize on a thesis carried out at Saft, which studied the life degradation trajectories of Li-ion batteries using Gaussian
processes. Building on this work, the student will model the life degradation of a battery that has not yet been studied, focusing on the transfer of
hyperparameters from existing models.
Study plan: Implementation of thesis work ([1], [2]) on the modeling of Li-ion battery life degradation based on Gaussian processes on a new battery
dataset. Exploration of context-sensitive domain adaptation methods.
Development of a life degradation model for a new type of battery based on this preliminary study.
Implementation of an evaluation process for the model thus developed.
References :
[1] B. Larvaron, M. Clausel, A. Bertoncello, S. Benjamin, G. Oppenheim, Chained Gaussian processes to estimate battery health degradation with
uncertainties. Journal of Energy Storage, 67, 107443, 2023
[2] B. Larvaron, M. Clausel, A. Bertoncello, S. Benjamin, G. Oppenheim, Chained Gaussian processes with derivative information to forecast battery
health degradation. Journal of Energy Storage, 65, 107180, 2023
[3] Zhou, Zihao, et al. "Few-Shot Cross Domain Battery Capacity Estimation." Adjunct Proceedings of the 2021 ACM International Joint Conference on
Pervasive and Ubiquitous Computing and Proceedings of the 2021 ACM International Symposium on Wearable Computers. 2021.
Keywords:
Machine learning, batteries, life cycle degradation estimation, domain adaptation
Conditions:
This internship, financed by SAFT, will take place at CRAN in Vandoeuvre-lès-Nancy (54) between April and September.
Expected profile: This is a Master's or engineering school internship for a student in data science, statistics, etc., but other profiles may be accepted. Python
skills desirable and good command of English expected.
Expected profile: This is a Master's or engineering school internship for a student in data science, statistics, etc., but other profiles may be accepted. Python
skills desirable and good command of English expected.
Department(s):
| Biology, Signals and Systems in Cancer and Neuroscience |
Funds:
Funded by SAFT