Hardware-Aware Fast-Charging Framework for EV Battery Management Systems: Bridging 3D Electro-Thermal Dynamics and Power Electronic Constraints

Fast-charging protocols for lithium-ion batteries are constrained by spatial thermal accumulation and hardware limitations. While high-fidelity 3D electro-thermal models capture critical localized hotspots, their prohibitive computational latency fundamentally renders them incompatible with iterative optimization algorithms. This study proposes a computationally tractable, hardware-aware Multi-Stage Constant Current (MSCC) optimization framework. A 1RC-3D Finite Element Method (FEM) model is coupled with a Multi-Objective Improved Grey Wolf Optimizer (MOIGWO). To overcome computational bottlenecks, Cholesky factorization reduces 3D FEM evaluations to sub-second execution times. Furthermore, deterministic transformation filters, Current Deadband and SOC Simplex Mapping, are directly embedded into the optimizer to ensure physically executable current profiles. Validated against a 4S2P Li-ion NMC module, the model achieves high accuracy with RMSEs of 0.057 V and 0.69 °C. Pareto front analysis reveals that under nominal highway operating conditions (27 °C, hconv = 8 W/(m2 ⋅ K)), the optimal MSCC strategy reduces the charging time by 17.5% (from 2400 s to 1980 s) compared to the 1C CC-CV baseline. Simultaneously, it prevents high-frequency micro-stepping, thereby mitigating transition stress on power converters, and strictly bounding the maximum pack temperature below the 45.0 °C safety limit. This methodology establishes a numerically constrained, simulation-based architecture for electric vehicle fast-charging systems, providing mathematical hardware compliance without relying on heuristic solution rejection.