A Novel Lithium‐Ion Battery Remaining Useful Life Prediction Method by Learning Wiener Process‐Based Degradation Model via Physics‐Informed Neural Network

This paper introduces Wiener‐informed neural network (WINN), an innovative physics‐informed neural network (PINN) framework designed for the accurate prediction of the remaining useful life (RUL) of lithium‐ion batteries (LIBs). Accurately predicting the RUL of LIBs is essential for ensuring battery system reliability, yet it is challenging due to the inherent stochasticity of the degradation patterns. Although Wiener process models effectively capture this randomness, estimating their parameters with conventional methods can be challenging, as it is often computationally intensive and prone to local minima. To overcome these challenges, this paper proposed a method for learning the Wiener process model parameters by integrating the Wiener process into the neural network. The proposed WINN effectively predicts and estimates the nonlinear degradation behavior, accurately capturing both linear and time‐varying drifts in battery degradation. WINNs incorporate governing Wiener process equations directly into the neural network loss function, ensuring adoption of Wiener process physical rules and providing accurate and physically realistic predictions. This hybrid methodology integrates data and physical principles, thereby improving model generalization in situations with limited data availability. WINN effectiveness is confirmed using two distinct datasets: an experimental dataset and the Oxford battery degradation dataset. Experimental findings show that WINN outperforms traditional estimation methods, making it a more effective option for health monitoring tasks. This study advances sustainable energy storage by precisely predicting battery lifespan, a critical component for ensuring the effectiveness, safety, and sustainability of LIBs.