Data-driven Diagnosis for High Temperature Polymer Electrolyte Membrane Fuel Cell - Fault Detection, Identification and Mitigation

Abstract:
High Temperature Polymer Electrolyte Membrane Fuel Cell (HT-PEMFC) shows great potential and prospects in automotive and stationary application thanks to its competitive efficiency, higher tolerance to impurities in the anode gas, stronger adaptability to reformed fuels, waste heat of higher quality and much more simple water management system when compared with traditional Low Temperature PEMFC, which is majorly due to its higher operation temperature by employing the phosphoric acid-doped polybenzimidazole (PBI) membrane. However, degradation in the long-time operation, especially those caused by unexpected faults, hinders its commercial application. Therefore, this project is designed to detect and identify the unexpected faults by online diagnosis based on data-driven methods and further to mitigate them via effective approaches. Firstly, the fault-oriented degradation mechanism shall be investigated utilizing in-situ characterization methods both to collect fault data for the model training in the further diagnosis step and to build a bridge between the degradation mechanism and the data-driven diagnosis to make the Machine Learning (ML) methods more interpretable. Secondly, Fault detection and identification will be conducted by data-driven methods based on the trained ML model, using the features extracted from the test data as inputs, employing specific ML algorithms as tools, and producing the fault type and fault level as outputs. As mentioned above, the feature extraction will refer to the degradation mechanism analysis to make the data-driven methods more reasonable and interpretable. Finally, the possible ways to mitigate or reverse the influence of faults on HT-PEMFC degradation are to be proposed. It is expected that the lifetime of HT-PEMFC could be prolonged through the real-time diagnosis and in-time warning of the major faults combined with the effective mitigation means.

Funding: China Scholarship Council (CSC)