A Real-Time Optimal Energy Management Model Using a Stochastic Predictive Control Framework and Reinforcement Learning for Electric Vehicles

The proposed research advances a stochastic Model Predictive Control (MPC) approach that employs Reinforcement Learning (RL) to manage energy management in Plug-in Hybrid Electric Vehicles (PHEVs). At the outset, the power distribution of each component in a PHEV is detailed. A convergent and efficient RL regulator is trained using the Q-learning algorithm based on the distribution of operating power across several driving phases. The Markov speed forecasting framework (multi-step) utilises an RL regulator within a stochastic MPC framework to determine the necessary power supply for the projected period. The results of the quantitative simulation show that the proposed method achieves a level of fuel economy similar to that of the stochastic dynamic programming method. The effectiveness of the charge monitoring across different reference paths demonstrates that the offered solution can be used in online applications that require a high rate of calculations.

Перевод пока недоступен