A Mixed-State Particle Filter for Remaining Useful Life Prediction of Lithium-Ion Batteries

作     者：Huang, Mengqi Du, Zhengyu Lu, Ruibo Wang, Xiaoji Peng, Changhong 

作者机构：School of Nuclear Science and Technology University of Science and Technology of China Anhui Hefei230027 China Science and Technology on Reactor System Design Technology Laboratory Nuclear Power Institute of China Sichuan Chengdu610213 China 

出 版 物：《SSRN》 

年 卷 期：2024年

核心收录：

主　　题：Lithium ion batteries 

摘      要：Lithium-ion batteries are extensively used in medical, transportation, industrial, and other applications due to rapid charging and discharging, constant power output, and light weight. Accurately predicting their remaining useful life (RUL) is essential for safe, reliable operation and timely maintenance. This paper presents a two-stage model based on symbolic regression and particle filter, aims to address the issues present in the current research on lithium-ion battery lifetime prediction. These issues include difficulties in early prediction, using a single degradation model, and the failure to consider the uncertainty of prediction. The proposed model constructs a fuzzy transition point between the two stages based on the degradation process similarity of history and operational batteries. Historical battery degradation data were employed to train the two-stage degradation model using symbolic regression. Parameter updating and state prediction of the degradation model were carried out using an improved mixed-state particle filter. This method can calculate the batteries RUL and their distribution. The performance was evaluated by analysing the CALCE and MIT datasets. The prediction error was less than 22 cycles, over 50% lower than the commonly used FCNN and LSTM timing prediction algorithms, demonstrating the feasibility and effectiveness of the method. © 2024, The Authors. All rights reserved.