摘要
针对锂离子动力电池在充放电过程中存在局部特征造成健康状态(State of health,SOH)难以精确评估的问题,提出了一种基于优化双向长短期记忆(Bidirectional longs hort-term memory,BiLSTM)神经网络的锂电池健康状态估计方法。从充放电数据提取具有局部特征的健康因子(Health features,HFs),采用核主成分分析法(Kernel principal component analysis,KPCA)对HFs进行融合和降维处理,形成融合的健康指标(Form principal health,FHI)。通过变分模态分解(Variational mode decomposition,VMD)方法对融合FHI进行多尺度分解,以捕捉不同时间尺度上电池状态的变化分量。利用Dropout-BiLSTM-Attention模型对代表全局变化趋势的分量模态函数IMF1(Intrinsic mode function 1,IMF1)的SOH进行估计。通过电池老化数据集进行验证,结果表明,提取的健康因子能够有效追踪锂电池的退化过程,此模型的预测精度远低于对比模型,均方根误差(Root mean square error,RMSE)最低为0.522%和平均绝对误差(Mean absolute error,MAE)最低为0.317%。提出的模型对电池局部特征的预测具有良好的泛化能力和鲁棒性。
Aiming at the problem that it is difficult to accurately evaluate the state of health(SOH)of lithium-ion power batteries due to local characteristics in the charge and discharge process,a novel approach based on optimized bidirectional longs hort-term memory(BiLSTM)was proposed.Health features(HFs)with local characteristics were extracted from the charge-discharge data,and Kernel principal component analysis(KPCA)was used to fuse and reduce the dimensionality of these HFs.Form integrated health indicators,FHI).The fusion FHI was decomposed at multiple scales by variational mode decomposition(VMD)to capture the varying components of the cell state on different time scales.The SOH of the component representing the global trend of change(IMF1,Intrinsic mode function 1)will be estimated using the Dropout-BiLSTM-Attention model.Validated by the battery aging data set,the results showed that the extract of health factor can effectively track the degradation process of lithium-ion batteries,and puts forward the model prediction accuracy is far lower than the contrast,the root mean square error(RMSE)0.522% minimum and mean absolute error(MAE)min.0.317%.The proposed model has good generalization ability and robustness for the prediction of local feature extraction.
作者
欧阳名三
赵俊婷
OUYANG Mingsan;ZHAO Junting(School of Electrical and Information Engineering,Anhui University of Science and Technology,Huainan 232001,China)
出处
《黑龙江大学自然科学学报》
2025年第3期349-361,共13页
Journal of Natural Science of Heilongjiang University
基金
国家自然科学基金资助项目(51874010)
安徽省重点研究发展计划项目(202004A07020043)。
