Cathodic modelling and control for PEMFC in variable altitude environment

Accurate modeling and control of the air supply system for proton exchange membrane fuel cells (PEMFCs) in variable altitude environments are essential for the efficient operation of fuel cell vehicles (FCVs) at high altitudes. To enhance the adaptability to the altitude variations, this paper for the first time proposes a transfer learning-based modeling method for air supply systems. Specifically, the method utilizes the identified state-space model data for feature learning and fine-tuning based on minimal experimental data, thereby circumventing the challenges associated with obtaining large-scale experimental datasets. Based on the proposed transfer learning model, a novel controller incorporating fuzzy logic and PID neural network (FL-PIDNN) is proposed for the air supply control across various altitudes and operating conditions. Experimental results reveal that the proposed transfer model predicts the gas supply pressure and flow rate from sea level to 4000-meter altitude accurately, with a mean error less than 1%. Moreover, comparative results between the FL-PIDNN controller and the widely-used controllers demonstrate remarkable improvement of control accuracy, reduced response time, and significant decreases in the control jitter.