Research On Fault Diagnosis And Application Of Proton Exchange Membrane Fuel Cell Based On Support Vector Data Description

Proton Exchange Membrane Fuel Cell（PEMFC）has a great application prospect in many fields,such as power generation and transportation,due to its high energy density,fast start-up,clean and zero-emission advantages.However,PEMFC is a nonlinear complex system with multi-physics coupling,and it is prone to flooding,drying and fuel starvation during operation.The occurrence of these failures will reduce the reliability and durability of PEMFC,thus restricting its commercial application process.Therefore,timely identification of various faults of fuel cells through reliable fault diagnosis strategies is of great significance to preventing fault propagation,mitigating the harmful effects,and improving the performance and life of PEMFC.In this thesis,the following researches are carried out for the PEMFC fault diagnosis based on a data-driven method:Firstly,the typical datasets of PEMFC under flooding,drying and fuel starvation fault conditions are obtained by experiments.The ability of data-driven PEMFC fault diagnosis methods to identify faults is highly dependent on the quality of the obtained data set.This thesis comprehensively analyzes the causes,hazards and mitigation methods of PEMFC water management failures and fuel starvation failures.On this basis,the experimental scheme for acquiring data sets of PEMFC under various working conditions is designed.In the experiment,the PEMFC is induced to suffer from flooding,drying and fuel starvation by changing the operating conditions of the stack,and the typical data of the fuel cell under each failure are collected.Secondly,the support vector data description（SVDD）algorithm is improved for PEMFC fault diagnosis.The existing SVDD methods tend to have low accuracy when used for PEMFC multi-type fault diagnosis.This thesis proposes an SVDD multi-class classification method based on probability membership and a dynamic radius SVDD strategy to improve the classification performance of the SVDD method.Then the improved SVDD method is used to diagnose and identify the PEMFC water management fault and fuel starvation fault.The results show that the proposed method has better category identification performance.Finally,a PEMFC online fault diagnosis software is developed.In order to improve the practical value of the PEMFC fault diagnosis method and meet the needs of fuel cell fault identification in daily experimental testing,this thesis develops a PEMFC online fault diagnosis software for the fuel cell test platform in our laboratory based on the PEMFC fault diagnosis method described above.Experiments have verified that the developed fault diagnosis software can diagnose PEMFC flooding,drying and fuel starvation in real-time,which can help testers to know the PEMFC operation status in time.