Computationally efficient model predictive control for electric vehicle battery thermal management

Model Predictive Control (MPC) stands out as a promising approach for Battery Thermal Management (BTM). However, due to large thermal inertia of batteries, achieving optimal control of BTM necessitates longer prediction horizons, leading to increased computational loads and uncertainties. This study aims to address these challenges by developing computationally efficient MPC solutions for BTM. Two approaches, Adaptive MPC (AMPC) and Hierarchical MPC (H-MPC), are proposed alongside the baseline Nonlinear MPC (NLMPC). NLMPC employs linearization at the nominal operating point and updates state parameters online, thereby simplifying model complexity. Through AMPC, both accumulation constraint violation and execution time are notably reduced, by up to 15% and 76%, respectively. On the other hand, H-MPC utilizes one NLMPC in the planning layer and another one in the operating layer. The optimal battery temperature from the MPC in the planning layer serves as a reference for the MPC in the operating layer. Notably, the two NLMPC in the two layers adopt different prediction horizons and sampling times. Simulation results demonstrate that H-MPC achieves significant reductions in BTM consumption and execution time, by up to 8% and 89%.