A novel fault prediction method based on convolutional neural network and long short-term memory with correlation coefficient for lithium-ion battery

Among various batteries, the lithium-ion battery is the most widely used battery type. The issue of battery safety has received considerable critical attention. If the fault cannot be detected on time, it can lead to an accident of thermal runaway or even explosion. Therefore, a novel fault prediction method based on convolutional neural network and long short-term memory (CNN-LSTM) with correlation coefficient is proposed to improve the accuracy of fault prediction. First, CNN is used to learn the local features correlations of battery voltage and extract the hierarchical features correlations. Second, LSTM is used to learn the long-term dependencies between data from local features, to complete the data prediction and to save the model. Then, the saved CNN-LSTM model is used to predict the data of each battery cell in the battery module. Finally, the correlation coefficient is used to analyze the predicted values and determine whether any battery has failed. Real data of lithium-ion batteries are used to verify the failure prediction method proposed in this paper. The results show that the CNN-LSTM model has higher prediction accuracy and reliability compared with the LSTM model and the Bidirectional LSTM (BiLSTM) model. It can be seen that the method proposed in this paper is expected to be applied to the battery management system of actual vehicles in the future. •The proposed method is based on CNN-LSTM for voltage prediction.•Correlation coefficient is introduced for fault diagnosis based on predicted voltages.•Combining the advantages of CNN and LSTM improves the accuracy of fault prediction.•The proposed method has high stability and accuracy for the different temperatures.