본문 바로가기

Deep Q-Network-Based Controller for Cabin Cooling System of Electric Vehicles

문서 유형
연구출판물 · 학술지 논문
등재
SCIEScopus
저널 순위
JCR Q2Scopus Q1
저널
IEEE ACCESS11 · pp.137274 · 2023
피인용
Web of Science 7Scopus 9
FWCI (Scopus)
0.56
JCR IF
JCR 2022 3.9
ISSN
2169-3536

저자

  1. 최완식제1저자Pusan Natl Univ, Sch Mech Engn, Busan 46241, South Korea
  2. 안창선교신저자Pusan Natl Univ, Sch Mech Engn, Busan 46241, South Korea

초록

The efficiency of the thermal management system of electric vehicles is important because the thermal management system requires a significant amount of electric energy. Therefore, controllers of the thermal management system should be designed considering the efficiency. This paper proposes a deep Q-network-based controller for the thermal management system in electric vehicles. The deep Q-networks were designed to control each actuator and the observation signals, the action signals, and the reward function were designed to achieve requirements. The controller regulates cabin and evaporator air temperature by adjusting the compressor and cooling fan speed while minimizing energy consumption and adhering to system constraints. Unlike previous studies, this design process considers practical implementation, including a high-fidelity plant model, essential constraint conditions, and multiple objectives. Test results show lower energy consumption and better temperature regulation performance than a heuristically designed rule-based controller. This method can optimize thermal management system performance in electric vehicles, which have increased complexity and number of thermal loads that conventional control methods cannot adequately address.

키워드

Air conditioningdeep learningelectric vehiclesQ learningreinforcement learningHVAC

식별자

Scopus EID
2-s2.0-85179812188
WoS UID
001123814200001