摘要
针对锂电池健康状态(SOH)估计与剩余寿命(RUL)预测问题,设计一种基于双向长短期记忆(BiLSTM)神经网络模型的预测方法。首先,提取美国国家航空航天局(NASA)锂电池的容量数据,将容量数据转为SOH数据并作为模型输入数据。其次,建立双层BiLSTM神经网络,使用加速自适应矩估计算法(Nadam)优化函数动态调整学习率。然后,通过BiLSTM神经网络模型分析锂电池数据,建立电池容量、SOH和RUL之间的联系。最后,全连接层输出电池SOH的估计曲线,从而预测其剩余寿命。通过NASA数据进行预测实验,BiLSTM神经网络的RUL预测误差稳定在3以内,SOH预测曲线的拟合度稳定在94.211%~95.839%,BiLSTM神经网络具有更高的鲁棒性和准确性。
Aiming at the problems of state of health(SOH)estimation and remaining useful life(RUL)prediction of lithium batteries,aprediction method based on bi-directional long short term memory(BiLSTM)neural network model is designed.Firstly,the capacity data of lithium battery of national aeronautics and space administration(NASA)is extracted,and the capacity data is converted into SOH data and used as the model input data.Secondly,a two-layer BiLSTM neural network is built and the nesterov-accelerated adaptive moment estimation algorithm(nesterov-accelerated adaptive moment estimation,Nadam)is used optimization function to dynamically adjust the learning rate.Then,the lithium battery data is analyzed through the BiLSTM neural network model to establish the connection between battery capacity,SOH and RUL.Finally,the fully connected layer outputs the estimated curve of the battery SOH to predict its remaining life.In the prediction experiments with NASA data,the RUL prediction error of the BiLSTM neural network is stable within 3,and the fit of the SOH prediction curve is stable at 94.211%~95.839%.The BiLSTM neural network has higher robustness and accuracy.
作者
王义
刘欣
高德欣
Wang Yi;Liu Xin;Gao Dexin(School of Automation and Electronic Engineering,Qingdao University of Science and Technology,Qingdao 266061,China)
出处
《电子测量技术》
北大核心
2021年第20期1-5,共5页
Electronic Measurement Technology
基金
国家自然科学基金(61673357)
山东省重点研发计划项目(公益类)(2019GGX101012)
山东省高等学校科学技术计划项目(J18KA323)
山东省研究生导师指导能力提升项目(SDYY18092)资助。
关键词
锂电池
健康状态估
剩余寿命预测
双向长短期记忆
lithium battery
state of health
remaining useful life prediction
bi-directional long short-term memory