Lithium Battery Health State Prediction Based on LSTM

Battery State of Health (SOH) is a crucial parameter for characterizing battery performance, and accurate SOH estimation is of great significance for battery management and maintenance. Given the unique advantages of Long Short-Term Memory (LSTM) models in processing time series data, this paper proposed an improved LSTM network specifically for lithium batteries and compares it with traditional Back Propagation (BP) neural networks and Temporal Convolutional Network (TCN) to demonstrate its superiority. The paper begins by collecting real-time battery operation data and preprocessing it, proposing a method for calculating the Battery State of Health (SOH). Subsequently, average charging voltage, average charging current, average discharging voltage, and average discharging current are selected as feature indicators to predict the battery's SOH. By comparing and analyzing the true values with the predicted values of the three methods, as well as using evaluation metrics such as Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Relative Error (RE), it is found that the accuracy of the LSTM model is far exceeding the other two methods. Experimental results indicate that the SOH estimation algorithm based on LSTM exhibits both accuracy and feasibility.