锂离子电池分数阶模型的多参数在线辨识方法

Multi-parameter Online Identification Method for Fractional-order Model of Lithium-ion Battery

锂离子电池分数阶模型由于含有表征老化机理的参数,被用于电池老化研究并期望据此实现在线的电池老化过程探究。在电池的有限使用工况和产品级检测条件下实现对该模型尽可能多的老化参数的在线辨识,将有助于该模型的在线应用。基于此,提出1种基于反向传播神经网络的多参数在线辨识方法。首先通过分析典型工况下参数灵敏度确定在线辨识的参数集,然后基于电池老化规律设计网络以及网络训练算法以提高辨识速度和准确性,同时设计验证方法以保证辨识参数的收敛性,仿真和实验结果验证在线辨识方法的辨识速度和准确性。

The fractional-order model of lithium-ion battery is used for battery aging research because it contains parameters that characterize the aging mechanism,and it is expected to realize the exploration of online battery aging process.The online identification of as many aging parameters as possible will promote the online applications of this model under limited operating conditions and product-level testing conditions of the battery.In this paper,a multi-parameter online identification method based on back propagation neural network is proposed.First,the parameter set of online identification is determined by analyzing the parameter sensitivity under typical operating conditions.Then,the network and training algorithm are designed based on the battery aging law to improve the identification speed and accuracy.At the same time,a verification method is designed to ensure the convergence of identification parameters.Finally,simulation and experimental results verified the identification speed and accuracy of the proposed online identification method.