Efficient diagnosis of water management faults in polymer electrolyte membrane fuel cells using optimized multi-sine excitation signal and TimesNet

Abstract

Accurate fault diagnosis is crucial for enhancing the reliability of polymer electrolyte membrane fuel cells (PEMFCs), a promising clean energy technology. Traditional methods, like Electrochemical Impedance Spectroscopy (EIS), provide insights into electrochemical processes but are limited by high costs and long measurement times. Recently, alternating current (AC) voltage response signals have emerged as an effective alternative, capturing essential diagnostic information without requiring steady-state operation. Multi-sine excitation signals, which improve the depth of online diagnosis, often suffer from redundancy, noise, and increased computational load due to wide frequency ranges, affecting the accuracy and efficiency of measurements. Based on the AC voltage response, this study proposes a novel fault diagnosis framework for PEMFCs, by integrating the distribution of relaxation times (DRT) analysis with the TimesNet deep learning architecture. DRT analysis improves the selection of characteristic frequencies by isolating those related to key electrochemical processes. These frequencies guide the construction of targeted multi-sine excitation, improving signal quality and fault sensitivity. TimesNet converts one-dimensional AC voltage into two-dimensional representation capturing complex temporal patterns, enabling accurate fault diagnosis. Experimental results demonstrated the effectiveness of this framework on different PEMFC health conditions, achieving up to 99.4 % accuracy. This framework provides faster and more reliable PEMFC diagnosis.