Energy Management Strategy of Fuel Cell Vehicles Based on Reinforcement Learning and Traffic Information

Abstract:

In order to improve the fuel economy and durability of fuel cell vehicles （FCVs）， an energy management strategy （EMS） for FCVs based on reinforcement learning （RL） and traffic information was proposed. First， a powertrain model was built based on parameters of key components. Then， based on the characteristics of urban road conditions， a traffic model was built in VISSIM and the vehicle driving data and traffic data were extracted. The traffic data was then used as input to predict velocity by using the long short-term memory neural network. Finally， based on the RL algorithm and using the predicted velocity， acceleration， and the state of charge of battery as inputs， the fuel cell system power was used as the output to design the EMS. The simulation results show that the hydrogen consumption per hundred kilometers of the proposed strategy is only 1.27% different from that of the dynamic programming strategy， and fuel cell system average power fluctuation is reduced by 5.01%，effectively improving the fuel economy and durability of the vehicles.