Study On Modeling And Control Of A Direct Internal Reforming Solid Oxide Fuel Cell System

The fuel cell is a kind of new and developing clean power generation equipment with high efficiency. From the viewpoint of the energy conversion, the fuel cell, as well as the primary and the secondary power sources, directly converts the chemical energy into electrical energy. From the viewpoint of the practical application, the fuel cell has unique properties. The fuel and oxidant needed in fuel cell operation are continuously supplied by external devices instead of being stored within fuel cell itself. The direct internal reforming solid oxide fuel cell (DIR-SOFC) not only has the general advantages of fuel cells (high efficiency, low pollution, low noise, high energy density and high reliability), but also has high operating temperature and special electrode catalyst. The DIR-SOFC can thus be directly fueled with the natural gas, coal gas and other hydrocarbons, and does not need any external fuel reformer. Therefore, the SOFC configuration can be simplified. Further, to increase the total efficiency of the power system and reduce pollution to the environment, the DIR-SOFC can be combined with gas and steam turbines to recycle the fuel cell exhaust gas which contains a lot of unused fuel and heat (the overall efficiency can thus be more than 80%). Additionally, the SOFC system has a large power generation capacity and covers a wide range of applications.The work is part of the project"Intermediate temperature planar SOFC modularized generation system"supported by the National Hi-Tech Research and Development Program (863) of China and the project"Harmonious nonlinear control of fuel cell-gas turbine hybrid generation system"supported by the National Natural Science Foundation of China. The dissertation takes the DIR-SOFC system fueled with methane-steam gas mixture as the study object to develop the model, numerical simulation technique and control strategy of the power plant. Based on the kinetics of the methane steam-reforming and the mechanism model of the DIR-SOFC system, the steady and dynamic performances of the DIR-SOFC system under different operating conditions are detailedly analyzed and discussed. Based on the above work, the data-driven wavelet network (WN) methods (immune optimized WN (IOWN) and self recurrent WN (SRWN)) are applied to establish the input-output black-box model of the DIR-SOFC system. According to the analysis and discussion of the dynamic performance, the control schemes (fuzzy control and predictive control) suitable for the DIR-SOFC system are designed. The results obtained by different control methods are compared and analyzed. The main contributions and achievements of this dissertation are given below:1. The kinetic models of reactions in the methane steam-reforming process are established based on the kinetic theory of the methane steam-reforming. According to the models, the calculation modules of the equilibrium composition and reaction rates for the reforming process are built in MATLAB/SIMULINK simulation environment. The thermodynamic analysis of the reforming reactions and the simulation for the steady and dynamic behavior are performed based on the built calculation modules. The influences of important parameters (operating temperature, pressure, S/C, etc.) on the reforming process are investigated. The methane conversion, product equilibrium composition (methane, hydrogen, carbon monoxide, carbon dioxide) and reaction rates in the reforming process under different conditions are calculated. The simulation results are compared, and the states in reactions are analyzed and discussed.2. The mechanism model of the DIR-SOFC is established based on the kinetics of methane steam-reforming, principles of chemical energy-electrical energy conversion, and fundamentals of mass and energy transfer. The simultaneous differential equations are employed to describe the physical and chemical behaviors occurring in the DIR-SOFC. To obtain the solutions of the model equations, the computational fluid dynamics (CFD) method, combined with the need for research, is applied to deduce the two-dimensional matrix form of the DIR-SOFC model equations to perform the numerical calculation. The simulation codes (m files) are programmed in MATLAB, and masked blocks are generated so that they can be invoked in SIMULINK. Then the blocks are integrated together to build DIR-SOFC module which can facilitate manipulating and invoking. The influences of important variables and parameters (SOLID thickness, fuel flow, air flow, pressure, S/C and voltage) on operating states and performances of the cross-flow fuel cell are investigated. Dynamic processes of two-dimensional parameter distributions of fuel cell caused by step changes of inputs and parameters are presented, analyzed and discussed. The research results of DIR-SOFC performances show that: 1) At the fuel inlet, the temperature of DIR-SOFC reduces significantly. After the temperature reaches its minimum, it gradually rises along the anode flow channel and reaches its maximum in the middle of the anode channel near the cathode channel outlet. Then, the temperature gradually decreases along the channel. 2) Along the cathode channel, the cell temperature gradually rises and reaches its maximum near the cathode channel outlet. 3) At the anode and cathode inlet, the current density is relatively low. Along the anode channel, the current density gradually increases and reaches its maximum near the maximum-temperature point. Then, it gradually decreases. Along the cathode channel, the general trend of the current density is gradually increasing. Near the anode outlet, along the cathode channel, the current density gradually reduces because the fuel is almost consumed near this position. 4) Along the anode channel in the DIR-SOFC, the methane concentration gradually decreases, the hydrogen concentration first increases and then decreases, while the steam concentration first decreases and then increases. Along the cathode channel, the oxygen concentration gradually reduces and reaches its minimum near the maximum-current density point. 5) The increased SOLID thickness decreases the cell mean temperature, mean current density, fuel and oxygen utilizations. 6) The fuel flow rate decreased by 20% results in the decrease in cell mean temperature and mean current density. The fuel utilization accordingly rises and the oxygen utilization reduces. 7) The fuel flow rate increased by 20% results in the increase in mean current density, while the mean temperature reduces. The fuel utilization accordingly reduces, while the oxygen utilization gradually decreases following the sudden increase. 8) The air flow rate decreased by 20% results in the increase in cell mean temperature, mean current density, fuel and oxygen utilizations. 9) The air flow rate increased by 10% results in the decrease in cell mean temperature, mean current density, fuel and oxygen utilizations. 10) The pressure changed from 3bar to 2bar results in the decrease in cell mean temperature, mean current density, fuel and oxygen utilizations. 11) The pressure changed from 3bar to 5bar results in the increase in cell mean temperature, mean current density, fuel and oxygen utilizations. 12) S/C 2.14→1.8 results in the increase in cell mean temperature, mean current density, fuel and oxygen utilizations. 13) S/C 2.14→3.0 results in the decrease in cell mean temperature, mean current density, fuel and oxygen utilizations. 14) The cell voltage changed from 0.7V to 0.75V results in the decrease in cell mean temperature, mean current density, fuel and oxygen utilizations.3. Based on the WN modeling method, the multiobjective optimization immune algorithm (MOIA) is employed to optimize the network structure and initial state of the original WN to improve the accuracy of the WN model. In order to enhance the capability of model in learning from dynamic behavior of the DIR-SOFC and system, the improved SRWN is used. To implement the on-line model identification, the quadratic function with forgetting factor is taken as the objective function, and parameter updating formulas for the SRWN are deduced (including the updating formulas for dilation and translation parameters and iterative updating formulas for weights). Then, the convergence of the parameter updating algorithm is proved, and the adaptive law for the learning rates of parameters is proposed. Simulation results show that the IOWN and the SRWN (using objective function with forgetting factor) both succeed in modeling DIR-SOFC with relatively high accuracy. Further, the SRWN has better on-line learning ability.4. According to the conservation law of mass and energy and empirical formulas, models of components (combustor, compressor, turbine, heat exchanger) of DIR-SOFC system are established. The sub-modules for these component models are constructed in SIMULINK. Based on component modules, a 190kW DIR-SOFC generation system model is also constructed in SIMULINK. Several operating cases are set up to perform simulation experiments via the system model. Dynamic performances of the DIR-SOFC system under different conditions are obtained. Influences of different variables on system performances are analyzed. The manipulated variables (inlet fuel and air flow rates and added fuel flow rate for the combustor) are determined to control operating states and output performance of the generation system (stack temperature, fuel utilization and system output power). The typical intelligent control methods, fuzzy control and predictive control, are employed to design the control strategy for the DIR-SOFC generation system. Improved non-uniform membership functions for fuzzy variables, T-S fuzzy rules and weighted average defuzzification are applied in the fuzzy control algorithm. In the predictive control, the SRWN model is taken as the predictive model for the DIR-SOFC system. The simulation experiments on the DIR-SOFC system for on-line modeling and prediction are performed. The results show better on-line identification ability, adaptability and relatively high predictive accuracy of the SRWN model for the DIR-SOFC system. In order to show the advantage of the SRWN based predictive control method, the IOWN model is used as a substitute for the SRWN model in the predictive control. The simulation experiment results for control show that the SRWN based predictive controller generates smaller overshoot and shorter setting time compared to the fuzzy controller and IOWN based predictive controller. Further, the SRWN based predictive controller can also track the set points of control variables with high precision in a speedy and stable manner.