Joint Prediction of Li-ion Battery State of Charge and State of Health Based on the DRSN-CW-LSTM Model

In this paper, we compare the effects of different deep learning network models applied to the joint estimation of State of Charge (SOC) and State of Health (SOH), and propose a joint estimation method of SOC and SOH for Li-ion batteries based on DRSN-CW-LSTM networks. This method is based on long-short-term memory (LSTM) and Deep Residual Shrinkage Networks with Channel-wise Thresholds (DRSN-CW). The LSTM network is used to predict the SOH value of the Lithium battery, and the coupling process between the predicted SOH value and the SOC value in the data is performed. Then we use the data information of Lithium battery voltage, current, temperature, and capacity in the deep residual shrinkage network for feature extraction, and further fit the time series data trend by LSTM to achieve the prediction of SOC of Lithium battery in the use cycle. The adaptive noise data processing function can be implemented in the residual shrinkage module of the DRSN-CW network to eliminate the negative impact of lithium-ion battery data stream quality on SOC prediction to improve SOC estimation accuracy. This paper trains our proposed network using public datasets of lithium batteries at different temperatures and operating conditions, respectively, and also compares the prediction effects of three neural network models on two public datasets of lithium-ion batteries. The experimental results show that the MSE and MAE of the deep learning model proposed in this paper are controlled below 5% on both public datasets, and have better noise immunity and prediction performance with high estimation accuracy compared with the other three deep learning models.