Large-Language-Model-Enabled Health Management for Internet of Batteries in Electric Vehicles
Hui Peng, Chenyuan Liu, Heng Li
Abstract
Machine learning models have become a prominent technique for predicting battery state in electric vehicles (EVs). However, due to the significant variability in the operating environments and conditions of different EVs, traditional machine learning models often exhibit limited generalization capabilities. Additionally, the computational limitations of on-board chips in conventional battery management systems (BMSs) can lead to considerable computational overhead. Furthermore, acquiring large-scale battery data for model training can be economically prohibitive. To address these challenges, this article proposes an Internet of Batteries (IoB) approach for battery health management, leveraging large language model (LLM) to monitor battery health. First, this article introduces the concept of IoB in EVs. Subsequently, an experimental IoB system is established. Through comparisons with other machine learning methods, the study demonstrates that LLM exhibit strong generalization capabilities for predicting the battery data of EVs, even with small-scale data fine-tuning. The experimental results suggest that the combination of LLM and IoB may represent a promising advancement over traditional machine learning approaches.