Fault Prognostic Of Solid Oxide Fuel Cells

As currently energy crisis and environmental problems are becoming increasingly serious,the solid oxide fuel cell(SOFC)as a clean,efficient in respect of energy conversion and pollution-free power generation device,has attracted worldwide attention among researchers.However,the fuel cell system includes a plurality of external components such as a reformer,a combustion chamber,a compression chamber and because of the complexity of the SOFC system,it is likely to fail during the operation of the SOFC system.Once faults occur,it will lead to SOFC system's performance degradation and in some serious situation,it may even lead the whole fuel cell to failure,which has become a major reason to limit the development of SOFC technology.If we can diagnose SOFC faults and predict their remaining life in the early stages,SOFC equipment's service life could be prolonged by arranging the maintenance for it.The main contents of this thesis include:(1)SOFC Fault Diagnosis Scheme Based on Least Squares Support Vector Machine(LS-SVM).Anode oxygen poisoning and cathodic water vapor poisoning may occur due to the influence of chlorine gas impurity and water vapor accumulation in the cathode.In this paper,the LS-SVM model is used to diagnose these two kinds of faults of SOFC.This model can effectively diagnose the fuel cell's working condition,more specifically,whether it is in normal condition,anode chlorine poisoning or cathodic water vapor poisoning,and its diagnostic accuracy is 99%.(2)Fault prognostics of SOFC based on HSMM.HSMM-based forecasting framework can be divided into three parts: HSMM parameter training,equipment current health estimates,and residual life estimates.In this paper,we use the full-life voltage degradation data of two kinds of fault types,anode chlorine poisoning and cathodic water vapor,to get a HSMM which reflects the system degradation process in these two kinds of fault cases respectively.In the case of different poisoning,the voltage degradation data of the battery is put into HSMM,and the Forward-Backward algorithm is used to calculate the probability of the fuel cells' state of health in the event of the occurrence of the voltage data.According to the maximum likelihood criterion,the current fuel cell's health state is estimated.Depending on the current health status of the fuel cell,the remaining life of the fuel cell can be predicted by the state transitionmatrix of HSMM and the average duration of each state.Experiments show that HSMM has a better predictive effect on the similarity of poisoning.(3)SOFC Fault Prognostics Method Based on Hybrid Model.When the voltage degradation of the current fuel cell is not similar to the voltage degradation of the HSMM life-span training data,the prediction result of HSMM is poor.In order to solve this problem,this paper proposes a hybrid fault prediction method combining HSMM with empirical model.If the current operating voltage of the fuel cell is similar to the HSMM training voltage,the residual life of the fuel cell is predicted by HSMM;If the similarity of the two voltage is low,the empirical model is used to estimate the remaining operating life of the fuel cell.The empirical model is a model that reflects the degradation trend of the fuel cell according to the operational data of SOFC;Otherwise,the two models are combined to predict the remaining life of the SOFC according to the similarity size.The experimental results show that the method based on the hybrid model has better effect than the empirical model and HSMM alone.