Rule and Q-learning based Hybrid Energy Management for Electric Vehicle

Abstract

To allocate the power between lithium battery and ultracapacitor in electric vehicle optimally, a new energy management strategy (EMS) is proposed based on the power manage rules and Q Reinforcement learning (RQL) strategy. In this hybrid energy management strategy, when the car is in a braking state or the lithium battery and the ultracapacitor energy are not enough, the output power of the lithium battery and the ultracapacitor is directly allocated based on the rules. Otherwise, an energy management strategy based on RQL is adopted, and the output power of the lithium battery is obtained through $Q$ value table. The performances of the proposed strategy are tested on the simulated electric vehicle experimental platform by 3 typical driving conditions, i.e., ECE condition, UDDS condition and HWFET condition. The results show that in contrast to the rule-based control strategy, the proposed strategy can better adapt to real-time power allocation, reduce the use of lithium batteries, and improve the efficiency of the hybrid energy system.