| Summary: | 博士 === 國立嘉義大學 === 資訊工程學系研究所 === 103 === In this study, the computational intelligent based dynamic energy management for hybrid electric vehicles has been discussed. The reinforcement learning method (RL), the fusion method reinforcement learning with fuzzy inference system (RLFIS), and fuzzy Q-learning (FQL) are proposed to observe with the vehicle driving environments and to figure out the adequate motor driving power to meet the driving conditions.
To demonstrate the methods, two types of hybrid electric vehicles are selected for the experiments. First, the powertrain of the fuel cell hybrid electric vehicle (FCHEV) is driven by the powers from a fuel cell stack and a dedicated battery. Second, the powertrain of the human-electric hybrid bicycle (Pedelec) is driven mainly by human’s pedal force with the assisted force from battery powered electric motor.
For the FCHEV, the experimental results show that the proposed RL dynamic energy management largely improves the hydrogen fuel utilization comparing with the existent FIS method under the given driving cycle, but only traded off a few stability of the state of charge of the battery. For the Pedelec, the experimental results show that the proposed FQL assisted power management largely improves the comfortability and remains equally safety comparing with the existent proportional assisted power management method and the RL assisted power management method under the given riding condition.
|
|---|
