Title: Deep reinforcement learning-based multi-objective energy management for fuel cell heavy-duty trucks

Authors: Zhongwen Zhu; Dongying Liu; Weihai Jiang; Zirui Zhang; Cheng Li; Chuanlong Ji

Addresses: Key Laboratory of Advanced Manufacture Technology for Automobile Parts, Chongqing University of Technology, Ministry of Education, Chongqing, China; School of Automotive and Transportation, Hefei University of Technology, Hefei, China ' School of Automotive and Transportation, Hefei University of Technology, Hefei, China ' Key Laboratory of Advanced Manufacture Technology for Automobile Parts, Chongqing University of Technology, Ministry of Education, Chongqing, China; School of Automotive and Transportation, Hefei University of Technology, Hefei, China ' School of Automotive and Transportation, Hefei University of Technology, Hefei, China ' School of Automotive and Transportation, Hefei University of Technology, Hefei, China ' School of Automotive and Transportation, Hefei University of Technology, Hefei, China

Abstract: Fuel cell hybrid electric vehicles (FCHEVs) have emerged as a key research focus due to low carbon emissions, high energy efficiency, and low noise, yet optimising energy management strategies (EMS) remains challenging. This study targets fuel cell health and hydrogen-electric efficiency, leveraging neural networks and deep reinforcement learning (DRL) to improve EMS. A full-vehicle Simulink model for a fuel cell heavy-duty truck (FCHT) was established based on experimental data, coupled with a fuel cell lifetime degradation model. A multi-objective optimisation strategy was designed via the deep deterministic policy gradient (DDPG) algorithm. Offline comparisons and hardware-in-the-loop (HIL) experiments verified its superiority: 0.57% total operating cost error and 0.7% state of charge (SOC) deviation, significantly enhancing fuel cell lifespan and economy-health synergy. This work provides a novel approach for FCHT EMS optimisation, facilitating FCHEV commercialisation and the transportation sector's low-carbon transition.

Keywords: fuel cell heavy-duty truck; energy management strategy; fuel cell lifetime degradation model; deep reinforcement learning; DRL; hardware-in-the-loop; HIL.

DOI: 10.1504/IJPT.2026.151999

International Journal of Powertrains, 2026 Vol.15 No.1, pp.18 - 39

Received: 09 Sep 2025
Accepted: 01 Nov 2025
Published online: 02 Mar 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article