Lithium-Ion Health Prediction for Battery Systems Using Deep A-LSTM-CNN Approach

This study developed a battery health indicator specific to lithium-ion cells and a moving-window technique for evaluating battery life. The curve of the partial charge voltage was used as a measure of cellular health. The battery's remaining life was estimated by fitting a linear aging model to the capacity data over a sliding window; forecast uncertainty was generated via Monte Carlo simulation. Together with an existing battery management system for electric vehicles, the proposed approach created methodologies for forecasting remaining usable life and estimating capacity. The suggested approach consists primarily of three stages: preprocessing, feature selection, and model performance evaluation. Preprocessing with conventional PCA is frequently used. The proposed approach uses filter- and fusion-based strategies to select features. Models are trained using the A-CNN-LSTM. In every case, the suggested technique is found to be better to both CNN and LS TM.