Abstract:

A prediction model based on data fusion and long short term memory network (LSTM) is proposed to address the issue of low prediction accuracy in predicting the remaining service life of lithium batteries,which mostly uses a single decay feature.Spearman correlation coefficient method is used to select the decay features with high correlation from the open source data set of lithium battery in the laboratory of University of Maryland (CALCE) with the capacity ratio as the prediction target.Auto encoder (AE) is used to perform data fusion on the decay features.The remaining service life prediction model of AE-LSTM lithium battery is established and compared with the prediction model of RNN and LSTM.The example shows that the proposed method has high prediction accuracy and can meet the needs of practical engineering.

Key words: remaining service life of lithium battery, feature selection, feature fusion, recurrent neural network