Discharge Voltage Time Series Classification of Lithium-ion Cells Based on Deep Neural Networks

Cells among lithium-ion battery packs often exist inconsistency due to production variability in industrial process, which the inconsistency may lead to the reduction of the available capacity and premature failure of the battery packs. To relieve the inconsistency of cells among a battery pack and screen cells with similar electrochemical characteristics for a series-cell configured battery pack, we present a new method for lithium-ion cell selection using discharge voltage time series classification (DVTSC) based on deep neural networks (DNN). In addition, we selected 61177 cells with capacities between 2.06 Ah and 2.10 Ah as experimental samples. The samples were preprocessed to obtain a labeled training set for the DNN model, which the size of the labeled training set is 3000, including 2000 training samples and 1000 test samples. Then we compared two classical baseline methods and four deep neural networks methods for time series classification. The results show that FCN has the best performance on three metrics and can effectively screen abnormal cells in the discharge voltage time series on an unlabeled test set.