Testing And Fault Diagnosis Of Solid Oxide Fuel Cell System

SOFC is a new power supply technology,it takes advantage of the chemical reaction between hydrogen and carbon compounds to convert the chemical energy of fuel into electricity.It has the advantages of high efficiency,quiet,environmental protection and no pollution.SOFC is currently in rapid development.However,to achieve the commercialization of SOFC,it is necessary to make a stable working condition of it.Therefore,the testing and fault diagnosis of SOFC stack and peripheral systems has become a necessary part of the relevant study.Firstly,according to the chemical process and thermal dynamic process of each sub-component,a SOFC system model with a reformer is built in this thesis.And the four most common typical faults are simulated on the system model(fan mechanical loss,reformer leakage,electrode delamination and stack leakage).The validity and accuracy of the model are verified by comparing the data obtained from actual experiment with the data obtained from system model,respectively,in normal and fault situations.Moreover,the normal and fault situations of SOFC are studied in this thesis.Based on the study,the internal mechanism of system faults is revealed,which lays the foundation for fault diagnosis in the later stage.Secondly,by using the simulation data and the related characteristics,according to the phenomenon and characteristics of different faults,the fault tree analysis and the neural network fault diagnosis model are designed for different faults.For the mechanical loss of the fan and the leakage of the reformer,due to the unique characteristics and parameters,the fault tree analysis method is used to detect these faults.And when it is detected that the fan mechanical loss has occurred,a redundant control strategy will be used to control the air flow in the system.But for the heap failure,because of its strong coupling characteristics,a neural network algorithm with good identification function for nonlinear and strong coupling systems is used to detect and separate these faults.Finally,In order to detect the effect of neural network diagnosis model on real data and facilitate the test of stack,a 5kW SOFC stack test and fault diagnosis system is built.It greatly improves the reliability and stability of stack test,and saves the cost of experiment.Using this platform to test the performance of stack in an ideal environment,collecting the typical characteristics of the stack performance and the corresponding experimental data.The accuracy of the proposed neural network model for SOFC stack fault diagnosis is further verified by the data obtained in the actual experiment.