Energy Storage Science and Technology ›› 2021, Vol. 10 ›› Issue (4): 1432-1438.doi: 10.19799/j.cnki.2095-4239.2021.0109

• Energy Storage Test: Methods and Evaluation • Previous Articles     Next Articles

Capacity estimation of lithium-ion batteries based on Gaussian process regression and feature selection

Yunfei HAN(), Jia XIE, Tao CAI(), Shijie CHENG   

  1. State Key Laboratory of Advanced Electromagnetic Engineering and Technology, School of Electrical and Electronic Engineering, Huazhong University of Science and Technology, Wuhan 430074, Hubei, China
  • Received:2021-03-16 Revised:2021-03-18 Online:2021-07-05 Published:2021-06-25
  • Contact: Tao CAI E-mail:M201871373@hust.edu.cn;caitao@hust.edu.cn

Abstract:

Because of the complex degradation characteristics of Li-ion batteries, it is a continuing challenge to accurately predict the state of health and remaining useful life of batteries, which limits the development of consumer electronics, electric vehicles, and grid energy storage technologies. The degradation mechanism of the batteries is complex and coupled with each other. Therefore, it is difficult to use a model-based method for accurate modeling. This study proposes a data-driven capacity estimation method for Li-ion batteries. By analyzing the evolution pattern of the voltage-discharge capacity curve with cycle aging, features with an electrochemical concept are selected as the model input, and the capacity of Li-ion batteries can be predicted by the Gaussian process regression (GPR) model. The input features of the model can be obtained online, and the capacity of the battery can be estimated without a full charge-discharge. The experimental verification was completed with the data sets for LCO/graphite batteries and LFP/graphite batteries. The results show that the method has a good generalization ability and can accurately estimate the capacity of different types of batteries. The proposed method is compared to the GPR model with electrochemical impedance spectroscopy as the input, and the results indicate that the proposed method can obtain better estimation accuracy. This highlights that the appropriate selection of various features can significantly affect the performance of the data-driven model for Li-ion batteries and can provide a reference for battery state prediction and diagnosis.

Key words: lithium-ion battery, capacity estimation, Gaussian process regression, feature selection

CLC Number: