Adaptive Power Optimization and Control for Automobile PEMFC Air Management System Based on Fuzzy Q-Learning and Prescribed Tracking Performance

The net power of the automobile proton exchange membrane fuel cell (PEMFC) system evidently varies with air mass flow and pressure. To enhance net power across various operating conditions, this study introduces an adaptive control strategy integrating operation parameter optimization. Initially, the fuzzy Q-learning method is proposed to select suitable reference signals for oxygen excess ratio (OER) and pressure to maximize net power. Additionally, to coordinate the air compressor and flow valve, an adaptive neural network control scheme is suggested for the multiple-input-multiple-output (MIMO) air management system. Furthermore, a prescribed performance function is devised to ensure minimal overshoot and rapid convergence for OER and pressure tracking. Through a sequence of simulations and hardware-in-loop (HIL) tests, the proposed control strategy is shown to effectively optimize net power under diverse operating conditions, exhibiting superior transient and steady-state performance compared to conventional methods. Quantitative comparisons further demonstrate the superiority of reinforcement learning (RL) algorithms and adaptive control techniques in enhancing both power and control system performance.