Fast charging of Lithium-ion Batteries (LiBs) is fundamentally limited by lithium plating, a degradation mechanism that undermines safety and accelerates performance loss. While prior machine learning methods in battery modeling have primarily focused on improving voltage predictions, this work introduces a physics-enhanced data-driven charging framework that directly predicts plating risk and enforces safety constraints during charging. A Gaussian Process (GP) surrogate is trained on high-fidelity Doyle–Fuller–Newman (DFN) simulations to capture overpotential dynamics near the anode–separator interface, the region most prone to plating. This surrogate is embedded within Model Predictive Control (MPC) schemes based on the Single Particle Model with Electrolyte (SPMe), to enforce physically meaningful safety constraints in real-time during fast charging. Simulation studies demonstrate that the GP-augmented MPC reduces lithium plating risk by up to 89%, achieving nearly a tenfold reduction in peak overpotential violations and a 95% decrease in cumulative degradation compared to standard Constant-Current Constant-Voltage and nominal MPC strategies. Importantly, these improvements were realized with marginal increase in total charging time. Furthermore, the same GP surrogate is integrated with an Equivalent Circuit Model (ECM) achieving comparable safety improvements with a 97% reduction in computational overhead, making it viable for embedded deployment. Overall, the proposed approach offers a modular, data-efficient and computationally tractable pathway toward safer and faster charging protocols, supporting the reliable adoption of electrified transportation.
扫码关注我们
求助内容:
应助结果提醒方式:
