基于粒子滤波及其改进算法的锂离子电池寿命预测

发布时间：2018-09-10 20:36

【摘要】：随着锂离子电池在便携式电子设备、电动自行车、电动汽车、人造卫星、航空航天等领域的广泛应用，其引发的故障将导致事故和巨大的经济损失，甚至严重威胁着人员的生命安全。因此，锂离子电池寿命预测这一问题的研究，，能切实保障实际工程中设备或系统的安全运行。粒子滤波算法作为一种基于模型的预测方法，在实际工程中得到了广泛的应用，同时在其不断发展的过程中也碰到一些有待解决的问题。本文基于统计学习理论，实现了一种粒子滤波与支持向量回归机相结合的方法，有效解决了单一基于粒子滤波的锂离子电池寿命预测中存在的问题。首先介绍了衡量电池寿命的指标，并从理论上分析了导致锂离子电池容量衰退的主要因素。通过Matlab曲线拟合工具箱计算出双指数形式的电池容量衰退模型参数，分别研究了有测量值更新的粒子滤波一步预测方法和无测量值更新的粒子滤波多步预测方法，并将其应用到锂离子电池寿命预测。然后，针对基于粒子滤波的锂离子电池寿命预测方法中所存在的不足，通过分析比较时间序列多步预测策略，实现了有测量值更新的支持向量回归机粒子滤波的一步预测算法和无测量值更新的支持向量回归机粒子滤波的迭代多步预测算法，并将上述融合预测算法对锂离子电池的寿命进行预测。通过仿真比较了支持向量回归机粒子滤波预测算法和粒子滤波预测算法的优劣性，发现支持向量回归机粒子滤波的预测方法不仅有效的提高了预测精度，而且还能实现对未来时刻系统状态的跟踪估计。论文最后总结了全文的主要工作和研究成果，并对今后的研究方向进行了展望。
[Abstract]:With the wide application of lithium ion battery in portable electronic equipment, electric bicycle, electric vehicle, artificial satellite, aerospace and so on, the failure caused by lithium ion battery will lead to accidents and huge economic losses. Even a serious threat to the safety of personnel. Therefore, the study of Li-ion battery life prediction can ensure the safe operation of equipment or system in practical engineering. Particle filter algorithm, as a model-based prediction method, has been widely used in practical engineering, at the same time, it has encountered some problems that need to be solved in the process of its continuous development. Based on the statistical learning theory, a method of combining particle filter and support vector regression machine is implemented in this paper, which effectively solves the problem of single Li-ion battery life prediction based on particle filter. First, the index of measuring battery life is introduced, and the main factors that cause the capacity decline of lithium ion battery are analyzed theoretically. Through the Matlab curve fitting toolbox, the parameters of the dual exponential cell capacity decline model are calculated. The particle filter one-step prediction method with the measurement value update and the particle filter multi-step prediction method without the measurement value update are studied respectively. It is applied to predict the life of lithium ion battery. Then, aiming at the shortcomings of the method of Li-ion battery life prediction based on particle filter, the multi-step prediction strategy of time series is analyzed and compared. The one-step prediction algorithm of particle filter of support vector regression machine with updated measurement value and the iterative multi-step prediction algorithm of particle filter of support vector regression machine with no update of measurement value are implemented. The fusion prediction algorithm is used to predict the life of lithium ion battery. The advantages and disadvantages of particle filter prediction algorithm and particle filter prediction algorithm of support vector regression machine are compared by simulation. It is found that the particle filter prediction method of support vector regression machine not only improves the prediction accuracy effectively. Furthermore, the system state tracking and estimation can be realized in the future. Finally, the paper summarizes the main work and research results, and prospects for future research.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM912

【相似文献】