Self-Supervised Learning and Federated Learning for First-Life Batteries

The estimation of the state of health (SoH) in electric vehicle (EV) batteries is critical for ensuring their reliability, safety, and performance. Traditional methods for SoH estimation often rely on extensive labeled data or centralized data processing, which can be impractical and raise privacy concerns. This paper explores the integration of federated learning (FL) and self-supervised learning (SSL), and presents a potential framework to address these challenges. By leveraging FL's decentralized approach and SSL's ability to utilize unlabeled data, a robust framework for SoH estimation can be achieved that maintains data privacy while maximizing the utility of available information. This approach shows a 31% improvement in mean square error (MSE) of SoH estimation accuracy when tested in a simulation environment.