Abstract
Steam reforming solid oxide fuel cell (SOFC) systems are important devices to promote carbon neutralization and clean energy conversion. It is difficult to monitor system working conditions in real time due to the possible fusion fault degradation under high temperatures and the seal environment, so it is necessary to design an effective system multifault degradation assessment strategy for solid oxide fuel cell systems. Therefore, in this paper, a novel hybrid model is developed. The hybrid model is built to look for the system fault reason based on first principles, machine learning (radial basis function neural network), and a multimodal classification algorithm. Then, stack, key balance of plant components (afterburner, heat exchanger, and reformer), thermoelectric performance, and system efficiency are studied during the progress of the system experiment. The results show that the novel hybrid model can track well the system operation trend, and solid oxide fuel cell system working dynamic performance can be obtained. Furthermore, four fault types of solid oxide fuel cell systems are analyzed with thermoelectric parameters and energy conversion efficiency based on transition and fault stages, and two cases are also successful by using the built model to decouple the multifault degradation fusion. In addition, the solid oxide fuel cell multifault degradation fusion assessment method proposed in this paper can also be used in other fuel cell systems.