摘要
预测锂电池剩余使用寿命RUL(remaining useful life)可以提高电池供电系统的稳定性和安全性,从而明确故障的发生并及时做出响应。在预测过程中粒子滤波PF(particle filter)常用于在线辨识模型参数,但当PF在线辨识参数时易出现粒子贫化问题,需要大量粒子才能完成状态估计,这将会导致预测结果不准确。为了提高RUL预测的准确性,提出一种基于时间递归神经网络TRNN(time recurrent neural network)和萤火虫算法FA(firefly algorithm)优化PF融合的锂电池RUL预测方法。首先,由于TRNN的泛化能力优于经验模型,并且易于捕捉容量退化的长距离依赖问题,因此选用其模拟各种条件下的电池退化模型;其次,基于FA优化的PF技术对TRNN模型参数进行递归更新,使粒子群移动到高似然区域,从而减少PF的贫化;最后,选择不同条件下不同电池的实验数据进行验证和比较。结果表明,与传统方法相比,该方法具有更高的RUL预测精度。
The prediction of remaining useful life(RUL)of lithium batteries can improve the stability and safety of battery-powered systems,thereby clarifying the occurrence of faults and making a timely response.During the prediction,particle filter(PF)is usually used for the online identification of model parameters.However,the problem of particle depletion is prone to occur when PF identifies the parameters online,and a large number of particles are needed to complete the state estimation,which will lead to inaccurate prediction results.To improve the accuracy of RUL prediction,a lithium battery RUL prediction method based on the fusion of time recurrent neural network(TRNN)and firefly algorithm(FA)optimized PF is proposed.First,since TRNN has a better generalization capability than the empirical model and is easy to capture the long-distance dependence of capacity degradation,it is selected to simulate the battery degradation model under various conditions.Second,based on the FA optimized PF technology,the TRNN model parameters are updated recursively to make the particle swarm move into a high-likelihood area,thus reducing the depletion of PF.Finally,the experimental data of different batteries under different conditions are selected for verification and comparison,and results show that the proposed method has a higher RUL prediction accuracy than the conventional methods.
作者
徐波
雷敏
王钋
XU Bo;LEI Min;WANG Po(College of Electrical and Information Engineering,Hunan University of Technology,Zhuzhou 412007,China)
出处
《电源学报》
CSCD
北大核心
2023年第2期138-145,共8页
Journal of Power Supply
基金
湖南省省市联合基金资助项目(2020JJ6071)。
关键词
锂离子电池
剩余使用寿命
时间递归神经网络
萤火虫算法
粒子滤波
lithium-ion battery
remaining useful life(RUL)
time recurrent neural network(TRNN)
firefly algorithm(FA)
particle filter(PF)
