The real-time prognostics of lithium-ion batteries is a fundamental requirement for ensuring safety, reliability, and lifetime management in advanced energy storage systems. This paper proposes a physics-informed liquid neural network (PI-LNN) framework that explicitly integrates Wiener degradation dynamics into the predictive process. By embedding physical constraints into the adaptive structure of liquid neural networks, the proposed approach enables accurate modeling of nonlinear degradation trajectories while preserving robustness under varying operational conditions. Experiments conducted on multiple lithium-ion battery degradation datasets demonstrate that the PI-LNN achieves higher accuracy and stability than conventional deep learning architectures. These results highlight the promise of PI-LNN as a practical solution for real-time health monitoring and residual life prediction of lithium-ion batteries.

Lithium battery, real-time prognostics, liquid neural network, physics-informed