基于RVM-PF融合算法的锂离子电池剩余使用寿命预测

Prediction of Remaining Useful Life of Lithium-ion Battery Based on RVM-PF Algorithm

针对传统PF(粒子滤波)算法在锂离子电池RUL(剩余使用寿命)预测中出现的估计精度低、过于依赖电池经验模型等问题,提出一种RVM(相关向量机)算法与PF算法相融合的锂离子电池RUL预测方法。通过RVM算法提取电池容量数据的相关向量,同时利用RVM的回归能力拟合同型号电池容量衰减轨迹,基于衰减轨迹构建PF算法中的状态空间模型,预测当前工况下电池容量衰减趋势。最后,将传统PF算法和RVM-PF融合算法的预测性能进行对比。结果表明,所提出的融合算法具有状态跟踪拟合度高、预测精度高、长期预测能力好等特点,且融合算法不依赖电池经验模型,具有较强的通用性。

The traditional PF(particle filtering)algorithm is characterized by its poor estimation precision and overdependency on battery empirical model in RUL(remaining useful life)prediction of lithium-ion battery.This paper proposes an RUL prediction method that combines RVM(relevance vector machine)and PF algorithm.Relevant vectors of battery capacity data are extracted by RVM and the battery attenuation trajectory of the same type of battery is fitted by the regression algorithm of RVM.Based on the attenuation trajectory,the state space model of the particle filter algorithm in PF is established to predict battery capacity attenuation under the existing operating condition.Finally,the prediction performances of the traditional PF algorithm and the RVM-PF algorithm are compared.The results show that the proposed method has the advantages of high state tracking fitting degree,high prediction accuracy and great long-term prediction ability.The proposed method does not rely on the battery empirical model and is universal.