Volume 1: Energy Innovations: Accelerated Deployment, New Concepts, and Emerging Technologies

Big data driven Deep Learning algorithm based Lithium-ion battery SoC estimation method: A hybrid mode of C-BMS and V-BMS Shuangqi Li, Hongwen He, Jianwei Li, Hanxiao Wang

Abstract

Batteries are the bottleneck technology of electric vehicles (EVs), which hosts complex and hardly observable internal chemical reactions. This paper presents a big data-driven battery management method utilizing the deep learning algorithm, with the ability to work stably under dynamic conditions and whole battery life cycle. First, a Deep Belief Network-Extreme Learning Machine (DBN-ELM) algorithm-based battery model is established to extract the deep structure features of battery data, and in which the rain-flow cycle counting algorithm is used to reflect the battery degradation phenomenon. Next, to improve real-time performance of Battery Management System (BMS), a conjunction working mode between the Cloud-based BMS (C-BMS) and BMS in vehicles (V-BMS) is proposed, and a battery State of Charge (SoC) estimation method based on the interaction between C-BMS and V-BMS is also presented. Using the battery data to verify the model effectiveness and accuracy, the error of the battery SoC estimation is within 3%.

Keywords electric vehicle, battery energy storage, battery management system, big data, deep learning