Volume 2: Innovative Solutions for Energy Transitions: Part I

Reinforcement Learning–Based Energy Management Strategy for a Series-Parallel Hybrid Bus Han Xuefeng, He Hongwen, Wu Jingda, Peng Jiankun, Chen Rui

Abstract

An energy management strategy based on double deep Q-learning algorithm is proposed for a SeriesParallel Hybrid Bus. The models of powertrain configuration and its main components are first established. Subsequently, a rule-based energy management strategy will be proposed. The China typical urban driving cycle (CTUDC) is used to evaluate the fuel economy performance of the two strategies studied in this paper. The simulation result indicates that the energy management strategy based on reinforcement learning decreased the fuel consumption by 7.3% per 100km compared to rulebased strategy.

Keywords Series-Parallel Hybrid Bus; energy management; rule-based; double deep Q-learning