摘要
由于锂离子电池经循环使用、环境变化易出现材料老化和缩短使用寿命等问题,很大程度上影响电池健康状态,因此研究其掌握电池老化规律,从而预测其使用寿命有着积极的意义。为提高电池健康状态的预测精度,使用能够更好预测长序列的长短期记忆神经网络(LSTM),并针对其模型超参数调整困难、收敛速度慢等问题,提出将LSTM和麻雀搜索算法SSA相结合的SSA-LSTM算法,将LSTM模型参数作为SSA的参数优化目标来完成建模和预测。实验结果表明SSALSTM算法预测准确性较高,有较好的应用价值。
The material aging and service life shortening of lithium-ion batteries tend to occur due to over recycling or environmental change,which greatly affects the battery state of health(SOH).Thus,it is of great importance to predict its service life via research on SOH and the law of battery aging.To refine the prediction accuracy,Long-Short Term Memory(LSTM)is introduced.Simultaneously,the SSA-LSTM algorithm that combines LSTM with Sparrow Search Algorithm(SSA)is proposed to tackle the difficulties in super parameters adjustment and slow convergence speed.The parameters of the LSTM model are taken as the optimization objective of SSA to complete the modeling and prediction.Finally,the experimental results demonstrate that the SSA-LSTM algorithm provides a higher prediction accuracy and a better application value.
作者
郭玄
朱凯
GUO Xuan;ZHU Kai(College of Automobile and Traffic Engineering,Jiangsu Institute of Technology,Changzhou,Jiangsu Province,213001,China)
出处
《电池工业》
CAS
2021年第3期131-135,共5页
Chinese Battery Industry
作者简介
郭玄(1995-),男,湖南人,硕士研究生,研究方向:锂离子电池寿命预测。Email:gx1836226090@163.com;通讯作者:朱凯(1984-),男,江苏人,博士,硕士导师,研究方向:车辆控制、智能交通。Email:fatkyo@jsut.edu.cn
