A model-based diagnostic technique to enhance faults isolability in Solid Oxide Fuel Cell systems

This work illustrates an innovative diagnostic technique able to improve fault isolability in Solid Oxide Fuel Cell (SOFC) energy conversion systems. On-board sensor reduction may induce fault clustering and, thus, hinder univocal fault isolation. According to the proposed technique, isolated system component sub-models, fed with faulty inputs, can be used to solve this issue. These models provide a set of redundant residuals, which react only if the related component is under faulty state. The technique is characterized and tested in simulated environment on an SOFC Anode Off-Gas Recycling (AOGR) system. Hydrogen external leakage, fuel and air heat exchangers efficiency reduction and recirculation unit malfunction are addressed and implemented in the complete system model. This latter is used to simulate system variables in both nominal and faulty conditions and compute residuals for fault detection and isolation. The sub-models are then used to introduce further residuals, and their behaviour is investigated at different fault magnitudes. The analysis is firstly performed in an ideal case scenario, considering the fault isolability that can be theoretically achieved. Then, practical application of the diagnostic algorithm is discussed, considering quantitative residuals deviation and properly analysing the effects of feeding the sub-models with inputs provided by both faulty and nominal models. The achieved results confirm the capability of the proposed approach to univocally isolate the considered faults in all the investigated conditions. Moreover, the analysis of the real case scenario proved the proposed algorithm suitable also for real applications.