【作者】 陈跃华；

【导师】 曹广益；

【作者基本信息】 上海交通大学 ， 控制理论与控制工程， 2007， 博士

【摘要】 燃料电池是21世纪最重要的一种发电技术之一。它利用燃料的电化学反应来获得电能,从而突破了常规以燃料燃烧的热能进行发电的方式所必须受到的卡诺循环效率的限制,因此可以获得更高的发电效率。根据燃料电池的工作温度,通常可以将其分为“高温燃料电池”和“低温燃料电池”。由于高温燃料电池的排气温度较高,可以和微型燃气轮机联合组成一个协同工作的发电系统,这就是一种重要的新技术—“联合发电系统”,其总体发电效率可以达到60～70%,因此被视为分布式发电装置的最佳形式。目前国内外对高温燃料电池/微型燃气轮机联合发电系统的研究主要集中在两方面:一是研发电池部件所需的新材料、研发适合与燃料电池配合的微型燃气轮机,并对联合发电系统的结构和装配工艺进行优化或改进;二是根据反应机理建立联合发电系统的仿真模型,基于所建立的模型研究各项操作参数对整体系统性能的影响,以确定系统的最优参数,为系统实际运行提供指导。但关于联合发电系统的可控模型和控制算法的研究很少,为了保证系统稳定运行、提高系统发电性能、最终实现商业化,解决联合发电系统有关控制方面的问题势在必行。本课题是上海交通大学燃料电池研究所正在进行的“863工程”项目“10kW级天然气熔融碳酸盐燃料电池发电系统研究”、国家自然科学基金项目“燃料电池-燃气轮机混合动力系统非线性对象的协调控制”课题的一部分。首先,根据反应机理和守恒定律分别建立了微型燃气轮机和熔融碳酸盐燃料电池(MCFC)的仿真模型,并基于模型分析了在额定工况和部分工况下各操作参数和系统结构对联合发电系统性能的影响。基于性能分析设计了联合发电系统的控制方案,首先针对联合发电系统的重要温度参数分别设计了三种智能非线性控制器,并根据仿真结果对其进行了性能对比和分析。然后采用了基于实验数据建模的两种方法对联合发电系统的输出功率特性进行建模,即径向基函数(RBF)神经网络和最小二乘支持向量机(LS-SVM)建模方法。基于建立的两种模型分别设计了模糊神经网络控制器和基于改进型遗传算法优化的非线性预测控制器,并进行了仿真和比较。本论文的主要工作包括:1、建立了微型燃气轮机的仿真模型。深入研究了微型燃气轮机各个组成部件的工作机理,在MATLAB/SIMULINK仿真环境里,采用模块化的建模方法,根据理想气体状态方程、质量守恒、能量守恒定律、热动力学公式和电功率的转换关系分别建立了微型燃气轮机的核心部件的模型,并将部件模型组合得到了整个微型燃气轮机的仿真模型;基于该模型进行了微型燃气轮机的额定工况和变工况性能的研究,分析了影响微型燃气轮机性能的主要参数。2、建立了MCFC电堆的仿真模型。同样是在MATLAB/SIMULINK仿真环境里,首先根据MCFC内部的电化学反应机理、质量和能量守恒定律、理想气体状态方程等建立了MCFC在气流方向上的一维仿真模型,基于该模型分析了影响MCFC发电性能的重要参数;然后将该模型和已建立的微型燃气轮机模型相结合组成了联合发电系统的模型,基于该模型对联合发电系统的运行进行了仿真,分析了系统结构和各个重要参数对系统性能的影响,为后继的控制系统设计打下了基础。3、基于所建立的联合发电系统模型和性能分析的结果,设计了以简单适用为目标的联合发电系统的多回路控制方案,并针对温度参数这一重要被控对象分别设计了三种控制器,对三者的控制效果差异进行了分析。首先采用了工业上广泛使用的模糊控制器;然后设计了Elman神经网络自校正控制器,该控制器使用一种动态递归神经网络-改进型Elman神经网络作为控制对象的辨识器,通过自校正算法得到控制量,仿真中取得了较好的效果;最后设计了将两者的优点集于一体的模糊神经网络控制器,对三种控制器进行仿真的结果显示三者均能够将温度参数稳定控制到设定的最优值,其中模糊神经网络控制器取得了最好的性能,对控制效果的差异进行了分析。4、着重进行了联合发电系统的输出功率控制方案设计,不同于前面使用的SIMULINK仿真模型,采用了基于控制对象输入输出数据建模的方法。首先采用了RBF神经网络对联合发电系统的输出功率特性进行建模,仿真结果显示RBF神经网络具有较好的辨识数据的能力;然后又使用了最小二乘支持向量机建模方法,比较了两种建模方法的特点和效果,结果显示最小二乘支持向量机方法能取得较好的性能,并对产生这种结果的原因进行了详细分析;基于两种建立的模型,分别设计了模糊神经网络控制器和基于改进型遗传算法优化的非线性预测控制器,仿真结果给出了两种控制器的控制结果,两种控制器的效果都比较好,其中非线性预测控制器能够使控制对象跟随参考轨迹快速稳定地达到设定值,取得了更好一些的控制效果,最后对结果进行了相应的分析。更多还原

【Abstract】 Fuel cells have become one of the most important generation technologies. It can convert chemical energy directly to electricity without the limitation of carnot cycle; therefore, the efficiency of fuel cell can reach a high level. In general, according the operating temperature, fuel cells are classified as“high-temperature fuel cell”and“low-temperature fuel cell”. Due to the high potential energy from the exhaust of high-temperature fuel cell, combination with a micro gas turbine to form a combined system is attractive. The efficiency of“combined system”can exceed the sum of the two equipments, therefore this method is considered as the optimum distributed generation mode.At present, the main researches on high-temperature fuel cell / micro gas turbine combined generation system are concentrated in the following two aspects in domestic and foreign: one is the researches and developments on new material of fuel cells stack, researches and developments on micro gas turbine adapted to combined generation system and on the optimization and improving the structure of combined generation system and assemble techniques; the other is the establishments of mathematical models according to the interior mechanism of combined generation system, and study the effects of system operating parameters on performance of combined system based on these models. However, there are only a few researches on control strategy for combined generation system. In order to guarantee stability of combined system during the process of operation and improve the generation performance, solving the control problems of combined system becomes extremely urgent.The task is part of the national 863 scientific project item“Analysis and control strategy for 10 kW-scale molten carbonate fuel cells power generation system”and the national natural scientific project item“Harmonious nonlinear control of fuel cell-gas turbine hybrid generation system”being researched in the institute of fuel cell of Shanghai Jiaotong University. Firstly, according the reaction mechanism, the mechanism models of a micro gas turbine and a MCFC stack are established, and the effects of operating parameters on the combined system are analyzed; Secondly, three kinds of intelligent nonlinear controllers for controlling the important temperature parameters of combined system are designed, including fuzzy controller, Elman neural network auto-adjust controller and fuzzy neural network controller. The comparisons of control performance among these controllers are realized in simulation tests. Finally, two kinds of modeling approaches are presented to model the output power of the combined system: radial basis function neural network and least squares support vector machine. Based on these two models, fuzzy neural network controller and a novel nonlinear predictive controller based on improved genetic algorithm are proposed to guarantee the optimum operation of MCFC/MGT combined system, comparisons between the two controllers are given and the results are analyzed. The main contributions and achievements of this thesis are given below:1、Establish the dynamic model of a micro gas turbine. According to the operating mechanism of micro gas turbine and using the law of mass conservation, energy conservation, ideal gas laws and thermodynamics formula, the components models of a micro gas turbine are built in MATLAB/SIMULINK environment. Based on this model, the static and dynamic performances of the micro gas turbine are analyzed.2、Build up the dynamic model of a molten carbonate fuel cell stack. The reaction mechanism is investigated deeply and using the conservation law of mass and energy, and ideal gas law, the dynamic model of a MCFC is built. Whole combination system model are made up of the MCFC model and the micro gas turbine model. Based on the combination system model, the effects of some important operating parameters on the whole system are investigated and evaluated for optimization and control of the whole system.3、According to the analysis results, the control strategy for the whole combined system is designed, and then, three kinds of controllers are presented to control the MCFC operating temperature and turbine inlet temperature parameter, including traditional fuzzy controller, Elman neural network auto-adjust controller and fuzzy neural network controller. Many simulations tests are implemented to evaluate the control performance of the three controllers, test results reveal that the fuzzy neural network controller can provide the best control performance. Finally, the reasons are analyzed.4、Design the control strategy for controlling output power of the combined system. Unlike the mechanism model, two kinds of data models based on the input-output data of controlled object are established. One is the RBF neural network model, the other is the least squares support vector machine model. Simulation results reveal that the LS-SVM model possesses better identification performance. Based on these two models, fuzzy neural network controller and a novel nonlinear predictive controller based on improved genetic algorithm are proposed. These two controllers are compared with each other in simulation tests. The simulation results show the effectiveness and merits of both the controllers. However, nonlinear predictive controller has advantage in the control performance.更多还原