Toward Reliable Battery Life Prediction: A Hybrid Data-Driven Framework with Uncertainty Quantification

Accurately predicting battery life is essential for performance management and system safety. Due to the complexity and diversity of internal mechanisms in lithium-ion batteries, their nonlinear characteristics directly give rise to uncertainty in the battery degradation process. However, most existing prediction methods do not fully account for the uncertainty caused by various factors and only provide a point estimate finally. To address this issue, this paper proposes a new framework that combines Random Forest and Conformal Prediction to predict battery life and quantify the uncertainty of the results. This approach leverages the efficiency of Random Forest while enhancing computational robustness and reliability through conformal prediction. The method utilizes early degradation data to select relevant features. Based on this, high-importance feature combinations are selected, and a Random Forest model is used to obtain point estimates. Then, the Conformal Prediction method is introduced to quantify uncertainty and generate prediction intervals with confidence levels and sample-specific bounds. Furthermore, the proposed method is compared against existing uncertainty quantification approaches, with coverage evaluation conducted to enhance the credibility of the prediction results. This method offers a new perspective for the practical application of battery lifetime prediction. Integrating uncertainty quantification into lithium-ion battery research can improve the reliability of the results and support decision-making in practical applications.