基于深度学习和粒子群算法的锂离子电池核温估计方法  

Core Temperature Estimation Method for Lithium-ion Battery Based on Deep Learning Method with Particle Swarm Optimization

作　　者：李毅超 王楠 段彬[1] 康永哲 张承慧[1] LI Yichao;WANG Nan;DUAN Bin;KANG Yongzhe;ZHANG Chenghui(School of Control Science and Engineering,Shandong University,Jinan 250061)

机构地区：[1]山东大学控制科学与工程学院,济南250061

出　　处：《机械工程学报》2023年第22期69-78,共10页Journal of Mechanical Engineering

基　　金：国家自然科学基金资助项目(U1964207,U1764258,62133007,61821004,62203265)。

摘　　要：在锂离子电池充放电运行过程中,其核心温度直接反应电池状态,是重要的安全性能指标。然而,电池核心温度无法直接测量,必须研发精准的核心温度估计方法。以核心温度相关性较高的可测变量(电流、电压、环境温度和表面温度)作为输入,建立锂离子电池双向长短期记忆神经网络(Bi-directional long short-term memory,Bi-LSTM)的核心温度预测模型,并引入粒子群优化(Particle swarm optimization,PSO)算法完成智能参数寻优,提高Bi-LSTM模型的预测精度。试验结果表明,在不同的充放电工况下,与决策树法、随机森林法等估计方法相比,本方法能够在宽环境温度下实现锂离子电池核心温度的准确预测,核心温度估计精度最高。During the charging and discharging operation of lithium-ion batteries,its core temperature directly reflects the battery state,which is an important safety performance indicator.However,the core temperature of the battery cannot be measured directly,so an accurate core temperature estimation method must be developed.The core temperature prediction model of Bi-directional long short-term memory(Bi-LSTM)neural network for the lithium-ion battery is established by taking the measurable variables(current,voltage,ambient temperature and surface temperature)which have a high correlation with core temperature as the input.The particle swarm optimization(PSO)algorithm is introduced to optimize Bi-LSTM’s parameters and improve its prediction accuracy.The experimental results show that under different charge and discharge conditions,compared with the estimation methods such as decision tree method and random forest method,this method can accurately predict the core temperature of lithium-ion battery under wide ambient temperature,and the estimation accuracy of core temperature is the highest.

关 键 词：锂离子电池 核温估计 双向长短期记忆神经网络 粒子群优化算法 

分 类 号：TG912[金属学及工艺—钳工工艺]