Oscillation Diagnosis in A SOEC System Based on Transfer Entropy Analysis

Oscillations of operation variables such as temperature and pressure threaten the security and stability of SOEC systems. To eliminate oscillations, it is crucial to locate the oscillation source, which, however, is difficult for model-based methods due to the complex system structure. To fill this gap, this paper proposes a novel data-driven framework for oscillation source diagnosing of SOEC systems. The measured data are first preprocessed and decomposed into different bands using singular spectrum analysis (SSA), Savitzky-Golay filter, and maximal overlap discrete wavelet transform (MODWT), effectively separating oscillations with diverse waveforms and frequencies. Transfer entropy (TE) is then employed to analyze the causal relationships between operation variables within different frequency bands, enabling accurate identification of oscillation sources. The proposed framework was validated using a dynamic model of an SOEC system in Simulink, demonstrating its effectiveness in diagnosing oscillation faults. The results indicate that the method accurately identifies oscillation sources in different frequency bands, providing valuable insights into the system’s behavior and potential for mitigating performance degradation.