基于差分电压和ICS-Elman神经网络的锂离子电池剩余使用寿命预测方法

REMAINING USEFUL LIFE PREDICTION OF LITHIUM-ION BATTERIES BASED ON DIFFERENTIAL VOLTAGE AND ICS-ELMAN NEURAL NETWORKS

锂离子电池被广泛应用于支撑新能源并网设备中,其剩余使用寿命(RUL)预测对设备运维管理极为重要,该文提出一种基于差分电压和改进布谷鸟搜索算法(ICS)-Elman神经网络预测锂离子电池RUL的方法。首先,对电池内部的电化学反应和外部的数据特征进行分析,选取结合电池内外特征的差分电压曲线作为特征提取对象,在充电差分电压曲线和放电差分电压曲线中选取相关特征;其次,考虑电池容量再生现象,选取Elman神经网络作为电池容量预测模型;然后,为提高预测精度,考虑利用改进的布谷鸟搜索算法对网络的初始权值和阈值进行参数寻优,ICS算法以改进概率公式、增加扩散因子、混沌初始化3种方法对传统CS算法进行改进,最终形成ICS-Elman预测方法;最后,利用NASA数据集和自测数据集对ICS-Elman方法进行验证,对比分析CS-Elman、Elman方法,结果表明所构建的ICS-Elman方法能更准确有效地预测锂离子电池RUL。

Lithium-ion batteries are widely used in equipment supporting new energy grid connection,and their remaining useful life(RUL)prediction is very important for equipment operation and maintenance management.This paper presents a method for predicting the remaining service life of lithium-ion batteries based on differential voltage and improved cuckoo search algorithm(ICS)-Elman neural network.Firstly,the internal electrochemical reaction and external data characteristics of the battery were analyzed,and the differential voltage curve combined with the internal and external characteristics of the battery was selected as the feature extraction object,and the relevant features were selected from the charge differential voltage curve and discharge differential voltage curve.Considering the phenomenon of battery capacity regeneration,a battery capacity prediction model based on Elman neural networks is established.In order to improve the prediction accuracy,the improved cuckoo search algorithm is used to optimize the initial weights and thresholds of the network.The cuckoo search is improved by three methods:improving the probability formula,increasing the diffusion factor and chaos initialization to form the ICS-Elman prediction method.Finally,the ICS-Elman method is validated by using NASA dataset and self-test dataset.The results show that the ICS-Elman method can predict the RUL of lithium-ion battery more accurately and effectively compared with the CS-Elman and Elman models.