锂离子电池循环寿命的融合预测方法

A fusion prediction method of lithium-ion battery cycle-life

针对传统基于粒子滤波的锂离子电池剩余使用寿命预测方法的不足:过度依赖电池经验退化模型和模型输入变量单一的问题,提出了一种相关向量机、粒子滤波和自回归模型融合的锂离子电池剩余寿命预测的方法。通过相关向量机提取电池历史数据的退化趋势,构建趋势方程替换以往的电池经验退化模型,作为粒子滤波算法的状态转换方程。引入自回归模型的长期趋势预测值,替换观测值构建粒子滤波算法的观测方程。将3种方法相融合估计电池剩余寿命。实验结果表明:融合方法不仅预测精度高而且采用数据驱动的方法避免了构建复杂的电池机理退化模型,通用性强。

According to the problems of traditional lithium-ion battery remaining useful life （RUL） prediction method based on particle filter, such as excessive reliance on battery experience degradation model and the single input variable of the model, a fusion RUL estimation ap- proach for lithium-ion battery is proposed based on relevance vector machine （RVM）, particle filter （PF） and the autoregressive （AR） model. The degradation trend of battery historical data is extracted by RVM, and the trend equation is built to replace the battery experience degra- dation model, which is adopted as the state transition equation of the PF algorithm. Long-term trend prediction values of the AR model are used to replace the real values, and then the observation equation of the PF algorithm is constructed. Three methods are integrated to estimate the battery RUL. Experimental results show that the prediction precision of fusion method is high, and the proposed data driven approach is more common because it can avoid building the complex experience degradation model based on battery failure mechanism.