Model Predictive Control-Based Online Optimization of Hybrid Electric Heavy Trucks Energy Management

The control strategy of the energy management controller significantly influences the fuel consumption performance of hybrid heavy-duty trucks. To optimize the performance of the energy management controller, numerous advanced control methods have been proposed. Dynamic Programming (DP) can calculate the globally optimal distribution of engine torque and motor torque based on pre-acquired global speed cycle information to achieve energy-efficient driving. However, it requires substantial computational resources, making real-time online application infeasible. The Model Predictive Control (MPC)-based energy management algorithm can utilize speed in-formation within the predictive horizon to calculate the required torque and optimize its distribution. However, it can only achieve local optimization and cannot fully exploit the energy-saving potential of torque distribution between power sources. This paper leverages the powerful fitting capability of neural networks to enable the online application of the DP algorithm. By utilizing the suggested driving patterns output by the neural network, the MPC algorithm is guided in torque distribution, thereby effectively enhancing energy-saving performance. To evaluate the performance of the proposed algorithm, simulation validation was conducted. The proposed method was compared with the original MPC algorithm. Simulation results indicate that the proposed method achieves superior energy-saving effects while maintaining power performance.