基于GWO-LSTM-TCN混合模型的锂电池荷电状态估计研究  

SOC estimation of lithium battery based on GWO-LSTM-TCN hybrid model

作　　者：李豪磊 赵升[1] 谢喜龙 张正江[1] 李泉坊 LI Haolei;ZHAO Sheng;XIE Xilong;ZHANG Zhengjiang;LI Quanfang(College of Electrical and Electronic Engineering,Wenzhou University,Wenzhou Zhejiang 325035,China;Zhejiang Juchuang Smartech Co.,Ltd.,Wenzhou Zhejiang 325014,China)

机构地区：[1]温州大学电气与电子工程学院,浙江温州325035 [2]浙江聚创智能科技股份有限公司,浙江温州325014

出　　处：《电源技术》2024年第11期2195-2200,共6页Chinese Journal of Power Sources

基　　金：温州市揭榜挂帅科技项目(ZG2023049)。

摘　　要：针对锂电池荷电状态(SOC)具有非线性、时变特性而无法直接测量的问题,提出了一种基于灰狼优化算法混合模型的锂电池SOC估计方法,利用长短期记忆网络(LSTM)和时序卷积网络(TCN)挖掘SOC特征信息,构建锂电池电压、电流与SOC的映射网络模型,引入灰狼优化算法(GWO)确定网络模型最佳超参数,采用马里兰大学公开的INR 18650-20R数据集对SOC混合模型进行实验验证。结果表明,GWO-LSTM-TCN网络模型对锂电池荷电状态的估计精度相较于GWO-LSTM网络以及GWO-TCN网络能更好拟合锂电池电压、电流与SOC之间的非线性映射关系,具有较好的模型准确性和泛化能力。A hybrid model utilizing the Grey Wolf Optimization algorithm is proposed to address the issue of the non-linearity and time-varying characteristics of the state of charge of lithium batteries.The method utilizes long-short-term memory(LSTM)and temporal convolutional networks(TCN)to extract SOC feature information and construct a mapping network model between battery voltage,current,and SOC.The GWO algorithm is used to determine optimal network model parameters.The public INR 18650-20R dataset from Maryland University is used to test the hybrid model's performance.The results demonstrate that the GWO-LSTM-TCN network model has improved accuracy in estimating the SOC of lithium batteries compared to the GWO-LSTM and GWO-TCN networks.It provides a more optimal fit for the nonlinear mapping relationship between battery voltage,current,and SOC,exhibiting excellent model precision and generalization capability.

关 键 词：锂电池 荷电状态SOC估计 GWO-LSTM-TCN 混合模型 

分 类 号：TM912[电气工程—电力电子与电力传动]