128 / 2025-05-19 16:27:10

A Fault Diagnosis and Localization Method of Proton Exchange Membrane Fuel Cell Stacks Based on Magnetic Field

PEMFC stack,Fault diagnosis,Multiphysics simulation,Magnetic field

Special Sessions > SSL 18 Advanced modeling, simulation, state assessment, and prognosis techniques for energy storage systems

Traditional PEMFC stack fault diagnosis methods based on voltage, current and impedance signals suffer from high cost, diagnostic latency, and limited fault discriminability. Magnetic field offers a promising non-intrusive and real-time monitoring choice but lack clear correlation mechanisms with internal stack states, thus limiting diagnostic precision. Therefore, this work develops a multiphysics simulation model to systematically analyzes the correlation between internal current and external magnetic field distributions under various typical faults and verifies the magnetic field superposition principle, which provides a solid theoretical foundation for fault diagnosis based on magnetic field. Building on this, an enhanced dual-feature selection random forest (EDFS-RF) method is proposed for both fault identification and localization to overcome the intrusiveness and low accuracy in existing methods. Magnetic field datasets from both simulation and experiment are used to train EDFS-RF model, which incorporate a dual-feature selection strategy to extract discriminative and physically interpretable features. Results show this method achieves high accuracy of 96.57% on simulation, and 97.45% on experiment, and outperforms conventional methods. This study enables non-destructive, real-time monitoring of PEMFC stack state, offering a promising new technique for online fault diagnosis and localization.