锂电池在不同放电区间下的剩余寿命预测

Remaining useful life prediction of lithium battery under different discharge intervals

随着锂离子电池的广泛应用,需要现有剩余寿命预测模型适应实际使用工况。针对锂离子电池在循环过程中放电区间对容量衰减影响较大的现象,为解决基于放电性能映射关系建立的剩余寿命预测模型应用范围较窄,提高车用锂电池剩余寿命预测模型适用性能,提出使用经验模态分解将容量分解为波动与趋势分量,并通过建立差分移动自回归模型以及广义回归神经网络分别进行预测,获得锂离子电池剩余寿命。选取NASA和CACEL电池数据集对模型进行验证,并对比基于蚁狮优化的相关向量机的方法,实验结果表明:提出的模型相比蚁狮优化的相关向量机的方法,对容量衰退的跟踪误差平均降低50%,能够实现不同放电区间下的电池老化预测,适用性能好,对电池容量再生现象追踪准确。

With the widespread use of lithium batteries,the existing remaining life prediction models need to be adapted to the actual operating conditions.To address the phenomenon that the discharge interval has a large impact on the capacity decay of lithium batteries during cycling,in order to solve the narrow application range of the remaining life prediction model based on the discharge performance mapping relationship and improve the applicability of the remaining life prediction model for automotive lithium batteries,we propose to use empirical modal decomposition to decompose the capacity into fluctuating and trend components,and establish a differential moving autoregressive model and a generalized regression neural network The remaining life of lithium-ion batteries is obtained by building a differential moving autoregressive model and a generalized regression neural network.The model is validated on NASA and CACEL battery datasets and compared with the method based on the correlation vector machine with antlion optimization.The experimental results show that the proposed model reduces the tracking error of capacity decline by 50%on average compared with the method based on the correlation vector machine with antlion optimization,and is capable of predicting battery ageing under different discharge intervals with good applicability and accurate tracking of the battery capacity regeneration phenomenon.