基于改进Transformer的锂电池剩余寿命预测

Prediction of remaining life of lithium batteries based on improved transformer

锂离子电池的剩余使用寿命(RUL)的准确预测对于评估电池状态至关重要。为了准确预测RUL,本文基于Transformer网络设计了以一种全新的模型p-Transformer。针对原始数据存在噪声的问题,构建了预处理前置编码器。在编码器内,原始输入数据叠加噪声信号,并实现数据降维压缩。而后将得到的深层数据输入到Transformer网络中,以学习原始数据中有价值的信息和深层次的非线性特征,得到最终预测结果。该模型在两个公共数据集上进行了大量实验,并与一些基准方法进行了比较。结果表明,本文提出的模型在预测RUL方面具有更小的误差,准确性更高。

The research on the prediction of the remaining useful life(RUL)of lithium-ion batteries is crucial for evaluating battery status.In order to accurately predict RUL,this article designs a new model which is called p-Transformer based on the Transformer network.A preprocessing encoder was constructed to address the issue of noise in the raw data.In the encoder,the original input data is superimposed on a noisy signal and data dimensionality reduction compression is achieved.Then,the obtained deep data is input into the Transformer network to learn valuable information and deep nonlinear features from the original data,and the final prediction result is obtained.The model was extensively experimented on two common datasets and compared with some benchmark methods.The results indicate that the model proposed in this article has smaller errors and higher accuracy in predicting RUL.