基于VMD-GPR的锂离子电池健康状态预测研究

Battery health state prediction based on VMD-GPR

为了对动力锂离子电池进行更好维护和管理,本文提出一种基于变分模态分解和高斯过程回归的锂离子电池健康状态(State of Health,SOH)预测方法。首先利用变分模态分解方法对原始训练数据集进行分解,可以生成固定个数的子训练集;然后在每个子集上训练高斯过程回归模型;最后利用自适应提升算法对训练的多个GPR模型进行集成,得到最终的预测模型。算例分析结果表明,所提出的VMD-GPR方法在预测锂离子电池SOH时,具有较好的准确率和广阔的应用前景。

In order to better maintain and manage lithium-ion batteries,this paper proposes a StateofHealth(SOH)prediction method based on variational mode decomposition and gaussian process regression.First,the original training data set can be decomposed using variational mode decomposition method to generate a fixed number of sub-training sets.Then the Gaussian process regression model is trained on each subset.Finally,multiple GPR models of training are integrated with adaptive enhancement algorithm,and the final prediction model example analysis results show that the PROPOSED VMD-GPR method has good accuracy rate and broad application prospect in the prediction of LITHIUM ion battery SOH.