Fault Diagnosis Strategy for Proton Exchange Membrane Fuel Cells Based on PSO-BP Neural Network Using Particle Swarm Optimization Algorithm

This paper presents a fault diagnosis strategy that integrates model-based and data-driven approaches for a 115 kW proton exchange membrane fuel cell used in vehicles. First, a stack subsystem model was developed in the MATLAB/Simulink platform based on the working principles and structure of PEMFC, and validated with experimental data. Subsequently, faults in the air and hydrogen inlet pipelines were simulated, and the resulting fault data were subjected to preprocessing steps, including cleaning, normalization, and feature extraction, to enhance the efficiency of subsequent data processing. Finally, a BP neural network optimized by particle swarm optimization was employed to achieve fault tree-based classification diagnosis. Experimental results indicate that the diagnosis accuracy of the BP neural network reached 96.04%, with an additional accuracy improvement of approximately 2.4% after PSO optimization.