Online energy management strategy considering fuel cell fault for multi-stack fuel cell hybrid vehicle based on multi-agent reinforcement learning

Wenzhuo Shi, Yigeng Huangfu, Liangcai Xu and Shengzhao Pang

Applied Energy, 2022, vol. 328, issue C, No S030626192201491X

Abstract: Because of the high-power demand of fuel cell hybrid vehicles, a multi-stack fuel cell system (MFCS) composed of multiple low-power fuel cell stacks (FCSs) instead of a high-power one has become a satisfactory solution. This is due to the modularity of MFCS, which is more reliable and durable. However, the hybrid power system (MHPSS) of an MFCS hybrid electric vehicle possesses not only MFCS, but also batteries to improve the dynamic performance of MHPSS. Due to the difference in characteristics of MFCS and battery, the energy management strategy (EMS) of MHPSS is the key of ensuring its safe and efficient operation. However, most of the existing MHPSS EMSs are complicated in design and complex to compute online. Besides, they also do not consider the robustness of EMS, that is, EMS has the ability to guarantee the safe operation of MHPSS when the MFCS fails. To solve the above problems, this paper proposes an EMS based on independent Q-learning (IQL) which is an algorithm of multi-agent reinforcement learning to maintain battery state of charge (SOC) and minimize hydrogen consumption. The proposed EMS is not only simple in design, but also can guarantee the normal operation of the MHPSS when the MFCS fails, and can be transplanted to execute online on a microcontroller unit or a field programmable gate array. The various parts of the MHPSS model are first built, then the IQL strategy (IQLS) is trained offline in the established model environment, and finally, the IQLS is ported to the hardware-in-the-loop platform for validation. In order to verify the effectiveness of the proposed EMS, the solution of dimensionality-reduced DP is also used to calculate the fuel economy. Through the experimental verification under different initial SOC and driving cycles, it can be seen that the IQLS proposed in this paper can achieve the goal of maintaining battery SOC and minimizing hydrogen consumption, and also has good generalization ability and safe operation ability under fault conditions.

Keywords: Energy management strategy; Multi-stack fuel cell; Hybrid power system; Multi-agent reinforcement learning (search for similar items in EconPapers)
Date: 2022
References: View references in EconPapers View complete reference list from CitEc
Citations: View citations in EconPapers (5)

Downloads: (external link)
http://www.sciencedirect.com/science/article/pii/S030626192201491X
Full text for ScienceDirect subscribers only

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:eee:appene:v:328:y:2022:i:c:s030626192201491x

Ordering information: This journal article can be ordered from
http://www.elsevier.com/wps/find/journaldescription.cws_home/405891/bibliographic
http://www.elsevier. ... 405891/bibliographic

DOI: 10.1016/j.apenergy.2022.120234

Access Statistics for this article

Applied Energy is currently edited by J. Yan

More articles in Applied Energy from Elsevier
Bibliographic data for series maintained by Catherine Liu ().