Neural Network and Reinforcement Learning based Energy Management Strategy for Battery/Supercapacitor HEV

A novel energy management strategy based on neural network (NN) and reinforcement learning is proposed to split the power of lithium battery and supercapacitor in the hybrid energy vehicle (HEV), which can meet the power required by the vehicle driving, improve the battery lifetime. First, the NN state-action model is gained by using traditional energy management controller, it generates new data and combined with the original data to optimize the NN model. Then, the action sequences are generated randomly, which will be evaluated by NN model, here, only the first action is adopted. The next action is re-predicted after this action is executed, which reduces the cumulative error caused by the inaccurate model. Finally, four typical driving patterns are selected, and the comparison with Deep Q-network (DQN) method is carried out, which verifies the performance of energy management strategy based on NN and reinforcement learning.