基于LSTM-Attention的锂电池SoC预测  

SoC Prediction for Lithium Battery Based on LSTM-Attention

作　　者：蒋永辉[1] JIANG Yonghui(College of Information Science and Technology,Hainan Normal University,Haikou Hainan 571158,China)

机构地区：[1]海南师范大学信息科学技术学院,海南海口571158

出　　处：《信息与电脑》2023年第9期99-101,106,共4页Information & Computer

基　　金：海南省重点研发计划项目(项目编号:ZDYF2021GXJS023)。

摘　　要：电池荷电状态(Stateof Charge,SoC)对新能源汽车安全行驶非常重要。文章提出一种基于长短期记忆网络(Long Short-Term Memory,LSTM)和注意力机制(Attention)的锂电池SoC预测方法。该方法将电池放电电流、电压、温度等电池工作历史参数作为输入,通过在LSTM神经网络中训练,并在LSTM神经网络中引入注意力机制,聚焦对当前SoC影响更大的历史参数,实现更为精确的锂电池SoC预测。实验证明,本方法比LSTM神经网络具有更好的预测性能,有较好的应用前景。The State of Charge(SoC)of batteries is crucial for the safe operation of new energy vehicles.This paper proposes a prediction method of lithium battery SoC based on Long Short-Term Memory(LSTM)and Attention mechanism.This method takes the battery discharge current,voltage,temperature and other battery operating history parameters as input,trains in the LSTM neural network,and introduces attention mechanism into the LSTM neural network to focus on the historical parameters that have greater impact on the current SoC,achieve more accurate SoC prediction for lithium batteries.Experiments have shown that this method has better predictive performance than LSTM neural networks and has good application prospects.

关 键 词：电池荷电状态(SoC)预测 长短期记忆(LSTM)神经网络 注意力机制 新能源汽车 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]