Remaining useful life (RUL) prediction is the core part of battery management system. The capacity regeneration phenomenon during battery capacity degradation interferes with the accuracy of RUL prediction. Therefore, this paper proposes an interpretable scheme named VPA model for lithium-ion battery RUL prediction by integrating algorithms with sufficient mathematical support. Firstly, trend signal (TS) and capacity regeneration signal (CRS) are obtained from capacity degradation sequence by variational mode decomposition algorithm. Then, prediction of TS and CRS is implemented by particle filter model and autoregressive integrated moving average model, respectively, and the prediction results of TS and CRS are superimposed as capacity degradation forecast. Finally, RUL prediction is performed based on degradation prediction result and failure threshold. Experimental results in engineering data prove the effectiveness of the proposed scheme.

剩余使用寿命（RUL）预测是电池管理系统的核心部分。电池容量退化过程中的容量再生现象干扰了RUL预测的准确性。因此，本文通过将算法与足够的数学支持相结合，提出了一种可解释的方案，称为 VPA 模型用于锂离子电池 RUL 预测。首先，通过变分模态分解算法从容量退化序列中得到趋势信号（TS）和容量再生信号（CRS）。然后，分别通过粒子滤波模型和自回归积分移动平均模型对TS和CRS进行预测，并将TS和CRS的预测结果叠加作为容量退化预测。最后，基于退化预测结果和故障阈值执行 RUL 预测。工程数据的实验结果证明了所提方案的有效性。