A Q-learning fuzzy inference system based online energy management strategy for off-road hybrid electric vehicles

被引：37
作者：
Bo, Lin [1 ,2 ]
Han, Lijin [1 ,2 ,3 ,5 ]
Xiang, Changle [1 ,2 ]
Liu, Hui [1 ,2 ,3 ]
Ma, Tian [4 ]
机构：
[1] Beijing Inst Technol, Sch Mech Engn, Beijing 100081, Peoples R China
[2] Natl Key Lab Vehicular Transmiss, Beijing 100081, Peoples R China
[3] Beijing Inst Technol, Adv Technol Res Inst Jinan, Jinan 250307, Peoples R China
[4] China North Vehicle Res Inst, Beijing 100072, Peoples R China
[5] Beijing Inst Technol, Sch Mech Engn, Natl Key Lab Vehicular Transmiss, Beijing 100081, Peoples R China
来源：
ENERGY | 2022年 / 252卷
基金：
中国国家自然科学基金;
关键词：
Off -road vehicle; Energy management strategy; HEV; Q-learning fuzzy inference system (QLFIS);
D O I：
10.1016/j.energy.2022.123976
中图分类号：
O414.1 [热力学];
学科分类号：
摘要：
In this paper, a Q-learning fuzzy inference system (QLFIS)-based online control architecture is proposed and applied for the optimal control of off-road hybrid electric vehicles (HEVs) to achieve better dynamic performance, fuel economy and real-time performance. A dynamic model, including a hybrid system, vehicle dynamics and road model, is established to obtain the state feedback according to the current driving environment under command. The optimal control strategy and objective function are both constructed by an adaptive network fuzzy inference system (ANFIS) due to its strong approaching ability. The fuzzy rules and parameters are trained online through the Q-learning algorithm and gradient descent method. This control framework provides a new control idea for the control of off-road vehicles. Without knowing the driving cycle in advance, it achieves a good control effect for different driving environments through online data collection and training. The QLFIS-based control strategy is compared to dynamic programming (DP)-based and rule-based strategies based on two different off-road driving cycles through simulation. The simulation results show that the vehicle dynamic performance and fuel economy are improved with respect to the rule-based strategy, while the calculation time is greatly reduced compared to that of the DP-based strategy. (c) 2022 Elsevier Ltd. All rights reserved.
引用
收藏
页数：11
相关论文
共 26 条
[1]   Deep reinforcement learning based energy management for a hybrid electric vehicle [J].
Du, Guodong ;
Zou, Yuan ;
Zhang, Xudong ;
Liu, Teng ;
Wu, Jinlong ;
He, Dingbo .
ENERGY, 2020, 201 (201)
[2]  
GLORENNEC PY, 1994, PROCEEDINGS OF THE THIRD IEEE CONFERENCE ON FUZZY SYSTEMS - IEEE WORLD CONGRESS ON COMPUTATIONAL INTELLIGENCE, VOLS I-III, P474, DOI 10.1109/FUZZY.1994.343739
[3]  
Glorennec PY, 2002, IEEE INT C FUZZY SYS
[4]   Energy management strategy research on a hybrid power system by hardware-in-loop experiments [J].
He, Hongwen ;
Xiong, Rui ;
Zhao, Kai ;
Liu, Zhentong .
APPLIED ENERGY, 2013, 112 :1311-1317
[5]   Power distribution strategy of a dual-engine system for heavy-duty hybrid electric vehicles using dynamic programming [J].
Hu, Jiayi ;
Li, Jianqiu ;
Hu, Zunyan ;
Xu, Liangfei ;
Ouyang, Minggao .
ENERGY, 2021, 215
[6]  
Li B, 2019, THESIS BEIJING JIAOT
[7]   Energy management for hybrid energy storage system in electric vehicle: A cyber-physical system perspective [J].
Li, Shuangqi ;
He, Hongwen ;
Zhao, Pengfei .
ENERGY, 2021, 230
[8]   An ensemble learning velocity prediction-based energy management strategy for a plug-in hybrid electric vehicle considering driving pattern adaptive reference SOC [J].
Lin, Xinyou ;
Wu, Jiayun ;
Wei, Yimin .
ENERGY, 2021, 234
[9]   THE BEST APPROXIMATION TO C-2 FUNCTIONS AND ITS ERROR-BOUNDS USING REGULAR-CENTER GUASSIAN NETWORKS [J].
LIU, BF ;
SI, J .
IEEE TRANSACTIONS ON NEURAL NETWORKS, 1994, 5 (05) :845-847
[10]   Two-stage robust distribution system operation by coordinating electric vehicle aggregator charging and load curtailments [J].
Lu, Xi ;
Xia, Shiwei ;
Gu, Wei ;
Chan, Ka Wing ;
Shahidehpour, Mohammad .
ENERGY, 2021, 226
123