Study On Fault Diagnosis Of Fuel Cell System And Energy Management Method Of Hybrid Power System

Proton exchange membrane fuel cell(PEMFC)is a green and environmentally friendly power generation device that converts hydrogen energy into electrical energy through a series of electrochemical reactions.It has outstanding advantages such as clean,high efficiency,high reliability,and stable power generation efficiency.At present,PEMFC has been successfully applied in many fields such as new energy vehicles,trams,smart power stations,etc.However,short service life and fault-prone defects restrict the large-scale popularization of PEMFC.In view of the above problems,it is urgent to carry out research on PEMFC system fault diagnosis and hybrid power system energy management methods.The main work and research results of the thesis are as follows:(1)The topology of the PEMFC system was studied,and its subsystem integration design was carried out according to the characteristics of the power reactor,and the 100 k W PEMFC system was designed and developed.The air supply subsystem integrates air filter,air compressor,humidifier and other original components to realize the functions of air impurity filtration,rapid response of air fluid control,humidification without additional water replenishment,etc.;the hydrogen supply subsystem integrates break valve,hydrogen circulation pump and other original components to realize the functions of hydrogen impurity filtration,hydrogen recycling,reaction hydrogen purification,automatic separation of reaction water,etc.;the heat dissipation and cooling subsystem integrates heat exchanger,deionizer and other original components to realize the functions of coolant ion rate control and dual cooling circuit integration.The heat exchanger,deionizer and other original components can realize the ion rate control of coolant and integrated heat dissipation of dual cooling circuits.At the same time,the control strategy is developed for the system characteristics,realizing air supply flow pressure regulation;hydrogen supply flow pressure control and hydrogen cycle management;cooling system temperature and temperature difference regulation,developing multiple protection control logic,and realizing efficient start-stop control of the system under the premise of this protection logic.(2)A generalized regression neural network(GRNN)and t-distributed stochastic nearest neighbor embedding(t-SNE)algorithm is proposed for the fault classification problem of PEMFC systems.system fault diagnosis method.The method uses the t-SNE algorithm to process the fault data with high dimensionality and strong coupling,so as to reduce the data dimensionality and computation,and extract new low-dimensional fault feature data;the GRNN method is used to classify the fault feature data,which can quickly and accurately diagnose the fault category of the system;the combination of the two methods can effectively improve the correct identification rate and reduce the computation time.The simulation results show that the proposed method can quickly identify four health states,namely,normal state,high inlet coolant temperature,low air pressure,and low spray pump pressure,and the correct identification rate can reach 98.75%.The effectiveness of the proposed method is verified by comparing with Back Propagation Neural Network(BPNN).(3)A fault diagnosis method based on Isometric Feature Mapping(ISOMAP)and Gradient Boosting Decision Tree(GBDT)is proposed for the multi-classification problem of complex faults in high-power PEMFC systems.The method uses Random Forest(RF)feature transformation to transform the original sample features into a high-dimensional,sparse space;the Isomap algorithm is used to downscale the data while preserving the original features;and the GBDT model is used to classify the feature data to achieve multi-classification fault diagnosis of PEMFC system health status.The proposed method can effectively diagnose five health states of PEMFC system: normal state,high reactor inlet coolant temperature,low air pressure,low shower pump pressure and low hydrogen pressure,and the correct detection rate can reach 98.00%.Finally,by comparing with BPNN and GRNN,the proposed method is verified to have high classification accuracy.(4)A rule-based finite state machine and an optimization-based dynamic programming energy management method are designed for the fuel cell hybrid system energy management method;considering the fuel cell durability,system hydrogen consumption and the real-time energy management method,a double-cycle energy management method based on dynamic programming and finite state machine is proposed to realize the real-time management of the output power of fuel cell and supercapacitor.Based on the RT-LAB semi-physical experimental platform,two different operating conditions were tested.The experimental results show that the proposed energy management method has significantly lower hydrogen consumption compared with the finite state machine energy management method,and the fuel cell output power fluctuation is smaller compared with the dynamic programming energy management method.The proposed method has significant improvement in hydrogen consumption and fuel cell durability,and has a strong binding effect on the supercapacitor SOC.