Remaining Useful Life Prediction of Li-ion Batteries Using an Improved CNN-LSTM Model

Addressing the challenge of accurately predicting the Remaining Useful Life (RUL) of Lithium-ion batteries, a predictive model leveraging the integration of an enhanced Convolutional Neural Network and Long Short-Term Memory (CNN-LSTM) Network is proposed. Based on correlation analysis, health factors closely related to capacity decline are extracted from battery charge-discharge cycle data as the input of CNN-LSTM network, and the global optimization capability of mixed strategy sparrow search algorithm (MSSSA) is facilitated to optimize the hyperparameters to improve the prediction ability. Based on the NASA battery data set and CALCE dataset, the RUL prediction under varying starting points and inputs are predicted and subsequently contrasted against results from different existing algorithms. The experimental results reveal that the RMSE and MAE of this method are both less than 2%, illustrating superior prediction accuracy and robustness compared to alternative algorithms.