A battery life prediction method based on graph neural network assisted by deep reinforcement learning

With the wide utilization of new energy batteries, the accurate prediction of battery life has become an urgent technical issue that needs to be addressed. Studies have developed battery life prediction models based on correlation analysis for feature extraction. This study introduces a DQN-based graph structure generation method that further incorporates the correlations among early-cycle feature data. Based on the correlations, an improved GraphSAGE model is developed, incorporating early-stage feature expansion, a graph global attention mechanism, and relative position information encoding. This design enables the neural network to focus on the most critical parts of the features while enhancing their representational capabilities. Experimental results demonstrate that the proposed method outperforms existing data-driven approaches. On the MIT dataset, it achieves a root mean square error (RMSE) of 82.27, a mean absolute error (MAPE) of 0.08, and a coefficient of determination (R²[jls-end-space/]) of 0.961. Furthermore, validation on the HUST dataset confirms the model’s robustness in long-cycle scenarios, achieving an optimal RMSE of 185, MAPE of 0.083, and R² of 0.757.