Research On Modeling And Control Of Solid Oxide Fuel Cell Based On Neural Network

Solid oxide fuel cell(SOFC)is the fourth generation of energy power generation technology after hydropower,thermal power and atomic power.It has the advantages of clean and efficiency,wide fuel sources,quietness and long service life.It has been highly concerned by many scholars at home and abroad.SOFC belongs of high temperature fuel cell.On the one hand,the research on its substance mostly focuses on material selection and structural design,on the other hand,the research mostly focuses on Dynamic Modeling and performance control.SOFC is a complex system with multiple inputs and multiple outputs.Its performance is easily affected by the external environment.A controller with strong anti-interference must be designed to ensure that it works in an ideal state.This thesis mainly studies the modeling and control of SOFC by using neural network.Firstly,considering the reasonable assumptions and basic voltage loss(activation polarization voltage,concentration polarization voltage and ohmic polarization voltage),the dynamic model of SOFC is established combined with the ideal gas equation of state and the law of mass conservation.Then,according to the SOFC dynamic model,the basic architecture of BP neural network is designed.The simulation experimental data are used to fit to obtain the SOFC BP neural network model with better accuracy and suitable control.On this basis,genetic algorithm and immune genetic algorithm are used to optimize the SOFC BPNN model.Finally,fuzzy controller,neuro fuzzy controller and improved neuro fuzzy controller are designed.It can control the output voltage of SOFC to keep the output voltage stable.At the same time,the system response under step disturbance is studied,and the anti-interference capability of the system is analyzed.Matlab/Simulink simulation results show that the SOFC dynamic model established in this thesis can effectively reflect the relationship between hydrogen input mol/sar flow,oxygen input mol/sar flow,water vapor input mol/sar flow,temperature and output voltage.The SOFC BP neural network model established according to the SOFC dynamic model can effectively predict the output voltage of SOFC,and the SOFC BP neural network model optimized by genetic algorithm and immune genetic algorithm has higher accuracy and can more accurately reflect the actual situation of SOFC.The design of fuzzy control strategy and neural fuzzy control strategy,the improvement of neural fuzzy control strategy can effectively control the output voltage of SOFC,expectations for SOFC output voltage in a short period of time to achieve,has good response speed and precision,and improved neural fuzzy control system output voltage accuracy,response speed is faster.In terms of anti-interference capability,the three control strategies have good anti-interference capability.With the improved neural fuzzy control algorithm,by contrast,the output voltage of SOFC by input mol/sar flow rate,hydrogen oxygen input mol/sar flow rate,water vapor input minimum interference of mol/sar flow rate,temperature and so on,stronger anti-interference ability,and further verified the reliability of the models established in this thesis,and the validity of the algorithm is designed.