Residual Life Prediction of Lithium Battery Based on the Improved Vector Regression Algorithm

Lithium-ion batteries are extensively utilized due to their remarkable performance in the realm of electric transportation and energy storage. This paper presents an improved Support Vector Regression (SVR) algorithm using the Pelican Optimization Algorithm (POA) to address its hyperparameter selection issue. By addressing the difficult problem of hyperparameter selection in SVR, this study significantly improves the robustness and the accuracy of the model used for predicting battery life. This study validates the effectiveness of the proposed algorithm by using battery degradation data from the NASA dataset. The results show that when using the POA-SVR algorithm for Remaining Useful Life (RUL) predictions, the MAPE values remain below 0.7% and the RMSE values remain below 0.02, indicating higher prediction accuracy.