A Physics-Informed Parameter Transfer Framework Between DFN and NTGK Models for Lithium-Ion Cells

Physics-based electrochemical models such as the Doyle–Fuller–Newman (DFN) framework provide high predictive accuracy for lithium-ion batteries but are computationally intensive, limiting their applicability in large-scale and real-time simulations. Reduced-order models such as the Newman–Tiedemann–Gu–Kim (NTGK) model offer improved computational efficiency but typically require experimentally fitted parameters, restricting their scalability across chemistries and operating conditions. This work proposes a physics-informed parameter transfer framework in which NTGK model parameters are derived directly from experimentally validated DFN simulation outputs using a regression-based formulation, thereby reducing dependence on direct experimental parameterization. The approach is applied to LCO–graphite and NMC–graphite cells across multiple discharge rates. The DFN model shows good agreement with experimental data at low to moderate C-rates, with mean absolute errors (MAE) in the range of 20–35 mV at 0.5C. The NTGK model parameterized using DFN-generated synthetic data accurately reproduces the DFN voltage response, with model reduction MAE values as low as 4.5 mV for LCO and 7.17 mV for NMC cells under low-rate operating conditions. Validation against experimental data yields MAE values up to 74 mV for LCO cells and 98 mV for NMC cells at higher C-rates. The proposed framework establishes a direct and physically consistent mapping between high-fidelity electrochemical models and reduced-order representations, enabling scalable and computationally efficient battery simulations while minimizing reliance on extensive experimental parameterization. This approach provides a practical pathway for integrating electrochemical fidelity into system-level and multi-physics battery simulations.