基于LE-ELM的锂电池热过程时空建模方法  

LE-ELM-based spatiotemporal modeling method of lithium battery thermal process

作　　者：吕洲 何波 黄镇泽 梁志勇 LYU Zhou;HE Bo;HUANG Zhenze;LIANG Zhiyong(Guangzhou Hong Kong University of Science and Technology Co.,Ltd,Guangzhou 511458,Guangdong,China;Guangdong Testing Institute of Product Quality Supervision,Guangzhou 510670,Guangdong,China)

机构地区：[1]广州港科大技术有限公司,广东广州511458 [2]广东产品质量监督检测研究院,广东广州510670

出　　处：《储能科学与技术》2022年第10期3200-3208,共9页Energy Storage Science and Technology

摘　　要：锂电池管理系统对于锂电池的效率、寿命和安全至关重要,而电池管理系统对电池的控制、热管理和故障诊断等都需要依赖于准确的电池热过程模型。然而锂电池热过程属于一种具有强非线性特征的分布参数系统,电池内部的温度分布是时空耦合的,并且具有无限维的特性,使得建模存在很大的困难。针对上述问题,本工作提出了一种基于LE-ELM的锂离子电池热过程建模方法。首先使用基于拉普拉斯特征映射(laplacian eigenmaps,LE)的局部非线性降维方法构建空间基函数,以表征系统固有的非线性拓扑特征;利用所得的基函数进行时空分离,获得原始数据的低阶时序表达;然后用极限学习机(extreme learning machine,ELM)以时间系数和对应的电流电压输入信号来近似低阶时序模型。最后集成辨识出的ELM模型与空间基函数,通过时空综合重构出锂离子电池的全局时空模型。为验证算法的有效性,使用所提出的方法对三元软包锂电池热过程进行建模。The efficiency,life,and safety of lithium batteries are influenced by the lithium battery management system.The control,thermal management,and fault diagnosis of the battery management system depend on the accuracy of the battery’s thermal process model.However,the thermal process of the lithium battery is a distributed parameter system with strong nonlinear characteristics.The temperature distribution inside the battery is spatiotemporally coupled and has infinite dimension characteristics,making modeling extremely challenging.To solve these problems,a Laplacian eigenmaps-extreme learning machine(LE-ELM)-based spatiotemporal modeling method for the thermal process of lithium batteries is proposed.First,a local nonlinear dimension reduction method based on LE is developed to learn the spatial basis function to represent the inherent nonlinear topological feature of the original system.Second,a lowdimensional representation of the original data can be generated using time/space separation and the spatial basis functions.Then,ELM is used to approximate the low-order time-series model with the low-dimensional representation and the corresponding current and voltage input signals.Finally,the spatiotemporal temperature distribution can be reconstructed using time/space synthesis.A thermal process of ternary soft-pack lithium battery was modeled using our proposed algorithm to verify the proposed method.

关 键 词：分布参数系统 锂电池热过程 拉普拉斯特征映射 极限学习机 

分 类 号：TM912.9[电气工程—电力电子与电力传动]