Degradation - Conscious Power Management of Fuel Cell Systems Using Reinforcement Learning
Year: 2026
Type: conference-paper
Abstract: The limitation of large-scale deployment by proton exchange membrane fuel cells (PEMFCs) is considered to be their long-term durability, since most power management approaches focus only on short-term efficiency. In this paper, the author suggests a reinforcement learning-based reinforcement-based degradation-aware power management scheme of a hybrid PEMFC-battery system. The formulated problem is a degradation-aware Markov Decision Process that is solved to use a Deep Q-Network. According to simulation results, the proposed strategy lowers cumulative degradation by up to 32%, lowers hydrogen usage by 5.6% and increases the estimated fuel cell life by about 46% and still provides better power tracking results. These findings indicate that reinforcement learning can be used to control a fuel cell power throughout a lifetime in an effective way.
Institution: SR University
Language: English
Subfield: Automotive Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7203661238