本论文采用了 Transformer模型与多种深度学习模型的组合模型来预测电池的健康状态(SOH)和剩余使用寿命(RUL).在NASA公开数据集合上进行了测试,使用电流、电压和温度来预测SOH,使用电流、电阻和阻抗来预测RUL.该模型首先利用卷积神经网络(convolution neural network,CNN)提取输入数据的空间特征,然后使用 双向长短期记忆网络(bidirectional long short-term memory,BiLSTM)提取输入数据的时间序列变化规律,再利用Transformer模型的多头注意力机制和前馈网络学习输入数据的特征表示,最后通过注意力机制进一步选取输入数据的时空特征中的重要部分,以共同预测SOH和RUL.实验结果表明,该模型在测试数据上的SOH预测均方误差(root mean square error,RMSE)达到0.084 85,RUL预测的RMSE达到1.46,其效果均优于传统方法.因此,该深度学习模型能够有效地提高电池SOH和RUL的预测精度和稳定性.
Prediction of SOH and RUL for lithium battery based on Transformer combination model
This paper employs a combined model of the Transformer model and various deep learning models to predict the SOH(State of Health)and RUL(Remaining Useful Life)of a battery.The model was tested on NASA's public data set,using current,voltage,and temperature to predict SOH,and current,resistance,and impedance to predict RUL.The model first uses a Convolutional Neural Network(CNN)model to extract the spatial features of the input data,and then uses a Bidirectional Long Short Term Memory(BiLSTM)model to extract the time series variation pattern of the input data.Then,the Transformer model's multi head attention mechanism and feedforward network is used to learn the feature representation of the input data.Finally,the attention mechanism is used to further select important parts of the spatiotemporal features of the input data and jointly predict SOH and RUL.Experiments on test data show that the Root Mean Square Error(RMSE)of SOH prediction reaches 0.084 85,and the mean square error of RUL prediction reaches 1.46,both of which are better than traditional methods.
state of healthremaining useful lifeconvolution neural networkbidirectional long short-term memory
NETL
NSTL
万方数据
