Big data driven vehicle battery management method: A novel cyber-physical system perspective

Abstract

The establishment of an accurate battery model is of great significance to improve the reliability of electric vehicles (EVs). However, the battery is a complex electrochemical system with hardly observable and simulatable internal chemical reactions, and it is challenging to estimate the state of battery accurately. This paper proposes a novel flexible and reliable battery management method based on the battery big data platform and Cyber-Physical System (CPS) technology. First of all, to integrate the battery big data resources in the cloud, a Cyber-physical battery management framework is defined and served as the basic data platform for battery modeling issues. And to improve the quality of the collected battery data in the database, this work reports the first attempt to develop an adaptive data cleaning method for the cloud battery management issue. Furthermore, a deep learning algorithm-based feature extraction model, as well as a feature-oriented battery modeling method, is developed to mitigate the under-fitting problem and improve the accuracy of the cloud-based battery model. The actual operation data of electric buses is used to validate the proposed methodologies. The maximum data restoring error can be limited within 1.3% in the experiments, which indicates that the proposed data cleaning method is able to improve the cloud battery data quality effectively. Meanwhile, the maximum SoC estimation error in the proposed feature-oriented battery modeling method is within 2.47%, which highlights the effectiveness of the proposed method.