Hyperparameter-optimized CNN and CNN-LSTM for Predicting the Remaining Useful Life of Lithium-Ion Batteries

This paper introduces novel advancements in predicting the Remaining Useful Life (RUL) of Lithium-Ion Batteries (LIBs) using Convolutional Neural Network (CNN) and Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) models, fine-tuned with Bayesian Optimization. Our study presents three key novelties. First, the research is rooted in the utilization of a publicly available dataset comprising 124 LIB cells, ensuring enhanced model generalization, and capturing a diverse array of battery usage patterns. Second, Bayesian Optimization is employed to optimize the hyperparameters of both models, leading to enhanced predictive accuracy. Third, we perform a rigorous direct comparison between the CNN and CNN-LSTM models, demonstrating the superiority of the CNN-LSTM model in RUL prediction by approximately 0.89%. Additionally, this study sheds light on the interpretability of the CNN-LSTM model, providing valuable insights into factors influencing RUL estimation. Both models exhibit high precision, with Mean Absolute Error (MAE) values of 85.6365 and 84.8746 cycles, respectively. The outcomes underscore the practical significance of accurate RUL prediction in LIBs, benefiting Electric Vehicles, battery manufacturing, and efficient maintenance planning. Our research contributes to advancing RUL prediction for LIBs in electric vehicles and energy storage systems.