锂电池分数阶等效电路模型构建与贝叶斯信息准则模型评估

Fractional-order Equivalent Circuit Model of Lithium-ion Battery and Model Evaluation Method Based on Bayesian Information Criterion

摘要:

随着技术革新，锂离子电池在动力汽车领域和储能领域得到了广泛应用；为了更好地评估电池的性能和寿命，需要建立适当的等效电路模型；根据构建的不同动态模型，采用最小二乘法对各种模型进行了参数辨识；并引入了贝叶斯信息准则(BIC)，该准则常用于综合性统计模型评估，通过给模型中的参数数量增加一个惩罚项，避免参数数量过多导致过度拟合，从而让模型的选择更加合理；使用马里兰大学提供的不同温度下的锂电池数据集进行参数辨识，通过仿真实验得到仿真端电压值，继而得到辨识误差的方差，最后算出各模型的BIC值；实验结果表明，不同温度下分数阶二阶等效电路模型的BIC都值最低，因此综合来看最为理想。

Abstract:

With technological advancements, lithium-ion batteries have been widely used in the field of electric vehicles and energy storage. To better evaluate battery performance and lifespan, it is necessary to establish appropriate equivalent circuit models. Based on different dynamic models constructed, the least squares method was applied for parameter identification of various models. The Bayesian Information Criterion (BIC) was introduced, which is commonly used for comprehensive statistical model evaluation. By adding a penalty term for the number of parameters in the model, it prevents overfitting caused by an excessive number of parameters, thereby making model selection more reasonable. Using lithium battery datasets provided by the University of Maryland under different temperatures, parameter identification was performed. Simulation experiments yielded simulated terminal voltage values, followed by the variance of identification errors, and finally, the BIC values of each model were calculated. The experimental results show that the fractional second-order circuit model has the lowest BIC values under different temperatures, making it the most ideal overall.