联合变分模态分解和长短时记忆网络的锂离子电池健康状态估计

State-of-health Estimation of Lithium-ion Battery Combining Variational Mode Decomposition and Long Short-term Memory Network

准确估计和预测锂离子电池的健康状态SOH(state-of-health)对新能源领域的发展至关重要,因此提出1种基于变分模态分解VMD(variational mode decomposition)和长短时记忆LSTM(long short-term memory)网络的锂离子电池容量衰减预测模型。首先采用VMD方法将原始电池容量衰减序列分解成比较单一的固有模态分量IMF(intrinsic mode function)序列,然后应用LSTM对分解得到的一系列IMF分量进行训练预测,最后对各IMF分量的预测值进行有效集成得到电池容量衰减序列的最终预测结果。基于美国国家航天局NASA(National Aeronautics and Space Administration)锂离子电池数据集选取的4块电池的放电容量衰减序列进行实验对比分析,结果表明:相较于LSTM、BiLSTM、EMD-LSTM、EMD-BiLSTM及CEEMDAN-LSTM方法,所提方法可以明显降低序列的复杂度,减少各IMF分量的模态混叠现象,具有很高的预测精度,优于其他预测模型,预测的最大平均绝对误差不超过5%,均方根误差和平均绝对百分比误差控制在4%之内。

The accurate estimation and prediction of the state-of-health(SOH)of lithium-ion batteries is critical to the development in the new energy field.A lithium-ion battery capacity decay prediction model based on variational mode decomposition(VMD)and long short-term memory(LSTM)network is proposed.First,the VMD method is used to decompose the original battery capacity decay sequence into a single intrinsic mode function(IMF)sequence.Second,the LSTM model is used to perform training and prediction on a series of IMF components obtained by decomposition.Finally,the prediction result of the battery capacity decay sequence is obtained by effectively integrating the predicted values of each IMF component.Based on the comparative analysis of the discharge capacity decay sequence of four batteries selected from the NASA lithium-ion battery data set,results show that compared with the LSTM,BiLSTM,EMD-LSTM,EMD-BiLSTM and CEEMDAN-LSTM methods,the proposed method can significantly reduce the complexity of the sequence and reduce the modal aliasing phenomenon of each IMF component.It has a high prediction accuracy,which is better than those of other prediction models.The maximum mean absolute error of prediction is not higher than 5%,and the root mean squared error and mean absolute percentage error are both controlled within 4%.