Fault diagnosis and prognostic of solid oxide fuel cells

作者：Wu, XiaoJuan^*; Ye, Qianwen

来源：Journal of Power Sources, 2016, 321: 47-56.

DOI：10.1016/j.jpowsour.2016.04.080

• Least squares support vector machine (LS-SVM)
• Fault diagnosis
• Hidden semi-Markov model(HSMM)
• Fault prognostic
• Solid oxide fuel cells (SOFCs)
• Durability

One of the major hurdles for solid oxide fuel cell (SOFC) commercialization is poor long-term performance and durability. Accurate fault diagnostic and prognostic technologies are two important tools to improve SOFC durability. In literature, plenty of diagnosis techniques for SOFC systems have been successfully designed. However, no literature studies SOFC fault prognosis approaches. In this paper a unified fault diagnosis and prognosis strategy is presented to identify faults (anode poisoning, cathode humidification or normal) and predict the remaining useful life for SOFC systems. Using a squares support vector machine (LS-SVM) classifier, a diagnosis model is built to identify SOFC different types of faults. After fault detection, two hidden semi-Mark models (HSMMs) are respectively employed to estimate SOFC remaining useful life in the case of anode poisoning and cathode humidification. The simulation results show that the fault recognition rates with the LS-SVM model are at best 97%, and the predicted error of the remaining useful life is within +/- 20%.

• 出版日期2016-7-30
• 单位电子科技大学