A Novel Real-Time Estimation and Neural Control Method for Automotive PEMFC Air Compressor Based on Simultaneous Coordination of OER and Cathode Pressure

Proton exchange membrane fuel cells (PEMFCs) represent a promising clean energy option for automotive applications. Within the context of the highly interdependent nature of the PEMFC system, the interaction between airflow and pressure is crucial, as focusing on one factor alone can lead to system instability. In this paper, a novel air compressor control strategy is presented to effectively coordinate airflow and pressure within the cathode channel, ensuring stability under varying load conditions. First, a nonlinear dynamic model of the air supply system is established by matching the characteristics of key components with experimental data. Second, a model-based internal state observer using an embedded cubature Kalman filter is proposed, along with an adaptive process to enhance the robustness to model uncertainties. Finally, a neural network-based air compressor control strategy is developed to achieve simultaneous coordination of air flow and cathode pressure. To optimize the strategy's overall performance, an enhanced particle swarm algorithm is employed. Comparative analysis shows that the proposed strategy has state estimation effect with higher robustness to system information, reducing the root mean square error of oxygen excess ratio and pressure tracking to 44.02% and 61.91% of the traditional method.