Markov Decision Process for the Energy Management of Parallel Hybrid Vehicles

In order to study the energy management strategy of single-shaft parallel hybrid electric vehicles, dynamic equations of the power-train system were set up to analyze the non-aftereffect property of the required torque. To achieve the optimization objective of minimizing the oil consumption under a fixed battery capacity, Markov decision process was carried out to implement the torque allocation strategy, and the policy iteration algorithm was used to solve the Markov decision model for energy management. In addition, the Markov decision process for energy management was simulated under the condition of J1015 driving cycles and Kunming driving cycles, and executed on line. The results show that compared to the strategy based on dynamic programming, the energy management strategy based on Markov decision process, which can be implemented on line, can make the battery capacity change more smoothly, but is globally sub-optimal in oil consumption: with the new strategy, the oil consumption is increased by 1.32 L per 100 km in case of J1015 driving cycles and 1.59 L per 100 km in case of Kuming driving cycles.