Real-time Simulation Based Multi-physical Fuel Cell Modeling And System Test

As a very active research area,fuel cells are considered to be one of the major candidates for future clean society and renewable energy development solutions.A fuel cell is an energy conversion device that directly converts chemical energy stored in a fuel into electrical energy without generating harmful gases that pollute the environment.Therefore,in recent years,fuel cells have gradually attracted attention in China,focusing on the rapid development of applications in the field of automobile transportation.However,compared to the development of simple functional applications of fuel cells,the corresponding fuel cell testing techniques and model analysis techniques have not attracted sufficient research attention.Therefore,this thesis aims to propose a general fuel cell modeling method,while verifying the performance of the fuel cell system through real-time simulation and degradation prediction.First,based on proton exchange membrane fuel cells(PEMFC)and solid oxide fuel cells(SOFC),the corresponding low-dimensional and multi-dimensional models are proposed.For PEMFC,a one-dimensional model of multi-physical domain was established.The proposed model can describe the phenomena among the electrochemical,fluidic and thermal domains.At the same time,for the medium temperature reversible tubular solid oxide fuel cell,a twodimensional multi-physical model was established.The proposed model can be used to describe the operating of both solid oxide fuel cells and solid oxide electrolysis cells(SOEC).By considering the phenomena among electrochemical,fluidic and thermal domain,the model presented in the thesis can accurately describe the multi-physical effects of the internal operation of the fuel cell and the electrolysis cell over the entire operating range of the current and temperature.In addition,an iterative numerical solver is proposed to describe the twodimensional distribution of axial physical model quantities along the tubular solid oxide fuel cell.Since the SOFC can use syngas as fuels,a syngas fed SOFC model was also established.As an important part of the model,the co-oxidation of hydrogen and carbon monoxide was discussed in detail.Secondly,the model of the PEMFC was verified by experimentally measured data.At the same time,the reversible SOFC model in both electrolysis mode and fuel cell mode was experimentally verified under different fuel gas partial pressures,different working temperatures and different current densities.Since the syngas gas fuel cell model contains the co-oxidation phenomenon of hydrogen and carbon monoxide,it is verified by different experimental conditions under different reaction temperatures,fuel gas partial pressure and working current density.The validity of the model indicates that the developed model can be used in embedded applications such as real-time simulation,which can help to design and test control and online diagnostic strategies for fuel cell power generation systems in industrial applications.Nest,real-time simulation is very important for fuel cell online diagnostics and hardware-inthe-loop(HIL)testing.However,for SOFC models,real-time simulation is difficult to be realized due to the numerical stiffness of the multi-dimensional multi-physical model.Therefore,the thesis first analyzes the numerical stiffness of the SOFC model in detail.The commonly used ordinary differential equation(ODE)solver was then tested to verify if its stability meets real-time simulation requirements.Finally,two suitable stiff ODE solvers are proposed to execute the fuel cell model in real-time.The experimental results on the general embedded real-time simulation platform show that the model execution time can reach the requirements of real-time simulation.At the same time,the stability of the solver and the high model accuracy under strong rigidity have also been verified.Finally,since PEMFC is susceptible to hydrogen impurities and unsuitable operating conditions,a decrease in output characteristics of the fuel cell may occur over time.Therefore,the prediction of fuel cell degradation is critical to the reliability of PEMFC systems.In the last part of this thesis,a data-driven fuel cell aging prediction method is proposed based on grid long short-term memory(G-LSTM)cell based recurrent neural network(RNN).LSTM can effectively avoid the gradient exploding and vanishing problem compared with conventional RNN architecture,which makes it suitable for the prediction problem for long period.By paralleling and combining the LSTM cells,G-LSTM architecture can further optimize the prediction accuracy of the PEMFC performance degradation.The proposed prediction model is experimentally validated by three different types of PEMFC: 1.2 k W Nexa Ballard fuel cell,1 k W Proton Motor PM200 fuel cell and 25 k W Proton Motor PM200 fuel cell.The results indicate that the proposed G-LSTM network can predict the fuel cell degradation in a precise way.The proposed G-LSTM deep learning approach can be efficiently applied to predict and optimize the lifetime of fuel cell in transportation applications.