State of Health Estimation and Prediction for Lithium-ion Batteries Based on Transfer Learning

Abstract:
In terms of the rapid development of electric vehicles and artificial intelligence, the battery management system is expected to get into a new era, where data-driven management algorithms are required. Novel state of health estimation and prediction method based on the data-driven method would help diagnose the health status and forecast the remaining useful life of batteries accurately and effectively, which ensure the safe running and maximize the lifetime of battery systems. This project aims to develop the artificial intelligence-based state of health estimation and prediction model for batteries using data in the source domain, and then expand it to different applications with transfer learning, where a well-designed retraining (self-updating) and domain adaptation strategies would be developed to improve accuracy and reliability with only a few or none labelled data for model updating. In this case, the health status at current time and in the future are supposed to be well estimated and predicted. This project would support scientific research by publishing journal articles and doing research reports, which explores the novel research on battery aging analysis and artificial intelligence development on battery health prognostics. It would also help the development of the electric vehicle industry to design the key algorithms in the next generation of battery management systems regarding practical applications.

Funding: CROSBAT