A Learning-Powered Model Predictive Control for Hybrid Electric Vehicles with Real-World Driving Data

In this paper, the energy management strategy for hybrid electric vehicle by using large-scale historical traffic data to improve the energy efficiency is explored. Firstly, the modeling of HEV powertrain is built. Then, an optimization problem is formulated to minimize the fuel consumption under the constraints of driver’s demand torque and physical limitations of powertrain. To derive the optimal solution for real-time application, an economic model predictive control algorithm is employed, where the learning concept is introduced to identify the feedback controller. Finally, the verification of the proposed algorithm is conducted in a high-performance simulator.