Thermal management and fast charging optimization for lithium-ion battery energy storage systems: A data-driven nonlinear model predictive control approach

Fast charging capability is critical for maximizing energy throughput and operational flexibility in lithium-ion battery energy storage systems. This paper proposes a data-driven thermal management strategy combining Linear Parameter-Varying (LPV) modeling with Nonlinear Model Predictive Control (NMPC). Using LUT-based variable-current charging data acquired from physical battery-pack tests across eight ambient temperatures from 17°C to 40°C, ridge regression identifies thermal dynamics at discrete temperature breakpoints, forming an LPV model that adapts to varying conditions. The model is embedded in NMPC framework to optimize charging current via rolling-horizon optimization under representative-temperature, current, and current-rate constraints. Under purged rolling-origin validation on unseen continuous time/SOC windows, the Ridge-LPV model achieved an R2 of 0.9331 and an RMSE of 0.991 m°C for one-step temperature-increment prediction relative to the recorded representative battery-temperature signal. Closed-loop controller performance was evaluated through MATLAB R2022a/Simulink simulations. Under matched paired operating conditions, the proposed strategy reduced charging time by at least 6.97% relative to the LUT baseline while maintaining the specified 50°C representative-temperature constraint. The proposed approach provides a computationally efficient and interpretable solution for NMPC-oriented fast-charging studies in battery energy storage systems.