Data-driven prediction of battery failure for electric vehicles

• DOI: 10.1016/j.isci.2022.104172
• Publication Date: 2022-03-28T16:12:52Z
• Authors: Jingyuan Zhao; Heping Ling; Junbin Wang; Andrew F. Burke; Yubo Lian
• Keywords: Multiphysics

Despite great progress in battery safety modeling, accurately predicting the evolution of multiphysics systems is extremely challenging. The question on how to ensure billions automotive batteries during their lifetime remains unanswered. In this study, we overcome challenge by developing machine learning techniques based recorded data that are uploaded cloud. Using charging voltage and temperature curves from early cycles yet exhibit symptoms failure, apply data-driven models both predict classify sample health condition observational, empirical, physical, statistical understanding multiscale systems. best well-integrated achieve a verified classification accuracy 96.3% (exhibiting an increase 20.4% initial model) average misclassification test error 7.7%. Our findings highlight need for cloud-based artificial intelligence technology tailored robustly failure real-world applications.