A BRB-LSTM-Transformer-Based Model for Lithium-Ion Battery Remaining Useful Life Prediction

Under the dual challenges of traditional energy shortages and environmental degradation, lithium-ion batteries have become a core driving force for new energy technologies because of their high energy density and long lifespan. However, estimating the remaining useful life (RUL) of lithium-ion batteries faces multiple challenges, and accurately assessing battery health is vital for enhancing battery management systems (BMS). Recently, significant progress has been made in RUL prediction through deep learning and machine learning algorithms, but existing models still face issues with hyperparameter dependence, data noise, and complex metric handling. To solve these challenges, this paper set a hybrid prediction model that combines traditional machine learning with deep learning. First, using principal component analysis (PCA) for feature reduction, and a belief rule base (BRB) module is introduced to improve data interpretability. Next, LSTM and an improved Transformer network are employed for time series modeling, incorporating temporal embedding mechanisms to better capture degradation trends. Finally, particle swarm optimization (PSO) is employed to optimize hyperparameters, improving prediction accuracy. The experimental results demonstrate that the proposed model performs better than other models on the NASA lithium-ion battery dataset. Through ablation experiments, the key roles of each component in the prediction process were verified.