随着环境问题和能源危机的日益突出，质子交换膜燃料电池（PEMFC）发动机已成为各国竞相研究的焦点。鉴于燃料电池系统集成的不够成熟和昂贵的实验费用，运用仿真方法建立系统的数学模型，对于理解其工作机理、分析和预测系统性能、提高系统的效率和鲁棒性等都具有十分重要的意义。本文以国家“十五”863《电动汽车》重大专项燃料电池城市客车项目为依托，在质子交换膜燃料电池的工作机理、电堆和空气系统的稳态优化、水热管理系统的设计方法、燃料供应与回收系统动态台架的建立和阳极压力跟踪算法、系统的动态模型及多变量耦合控制等方面进行了系统的阐述和分析。主要的研究成果包含单电池、子系统和综合系统等多个层次的工作。本文围绕燃料电池系统的建模与控制方法进行研究，首先，本文综合大量已有的单电池机理模型，简化出可用于系统级分析的PEMFC机理模型。然后在子系统研究中结合前处理和神经网络技术，对空压机和涡轮MAP进行了非线性拟合和模化匹配，并采用实数型遗传算法优化分析了三种不同拓扑结构中电堆和空气系统间的平衡；基于完整的散热器和冷凝器的解析设计方法和膜增湿器数学模型，首次比较了阳极回收、系统冷却水膜增湿等三种阳极增湿方式的水热管理特性；设计搭建了国内首个喷射器动态测控台架，采用xPC Target建立了硬件在环平台，并基于台架的数学模型设计了LQG线性控制器和模糊神经网络控制器用于压力跟踪。最后，建立了空气系统和燃料供应与回收系统的综合动态模型，在此基础上分析了多变量间的耦合和系统的可观测性，分别设计了基于状态观测的LQG线性控制器和基于神经网络在线辨识的模型预测非线性自适应控制器。仿真分析表明，多变量耦合控制具有良好的动态响应。此外，还提出了基于Stateflow的有限状态机分层控制的思路。本文的工作基本上涵盖了车用纯氢燃料电池系统的主要问题，对于燃料电池系统的设计、系统集成和控制设计等均具有一定的指导意义。
With the rise of energy crisis and environment problem, proton exchange membrane fuel cell (PEMFC) engine has been the focused solution in the whole world. Due to the immaturity of system integration and highly experimental costs, it is very significant for modeling and simulation to understand fundamentals, analyze and predict system performance and improve the system efficiency and robustness.Supported by the “State 863 Program Fuel Cell Bus Project” for the National Tenth Five-year Plan, this dissertation systematically described and analyzed the working mechanisms in PEMFC, the trade-off between the stack and the air system, design methodology of the system-level water and thermal management, build of test and control bench of fuel recirculation and the algorithms of pressure tracking, system dynamic model, and linear and nonlinear adaptive multivariable control. The main achievements in this paper involve the multi-level research on single cell, subsystems and integrated system.Aiming at the modeling and control of PEM fuel cell systems, author firstly synthesized the ancestor’s work and reduced a one-dimensional analytical PEMFC model which can be applied in the system-level analysis. Then by combining the technologies of preprocessing and neural network, nonlinear MAP of compressors and turbine were interpolated and extrapolated well. Analogy theory was used to match between size of stack and air system. For three kinds of air systems, i.e. single-stage boosting, two-stage serial boost and compressor/expander, a real-code genetic algorithm was used to search the global optimum air flow rate and pressure. Based on the theory of compact heat exchangers, the design methodology of radiator and condenser is described analytically in this paper. With the distributed model of membrane humidifier, the performance of water and thermal management in three types of anode humidification including anode recycling, membrane humidification with system cooling water were analyzed. In addition, the first domestic test and control bench of fuel recirculation was designed and built, and xPC Target environment was utilized to set up the hardware-in-the-loop platform. Based on the mathematic model of the bench, a state-feedback linear controller and a fuzzy neural network adaptive controller were designed for the set-point pressure tracking. Finally, based on the integrated dynamic model of air supply system and fuel recirculation system, the couples between multi variables and system observability were analyzed in the frequency domain. Then a two-freedom observer-based LQG controller and a nonlinear adaptive model predictive controller (MPC) with an on-line neural network identifier were designed. With the multivariable control algorithm, the close-loop transient response and robustness was satisfied. Moreover, the idea of hierarchical control based on finite-state mechanism was realized in MATLAB/Stateflow environment.The work in this paper basically dealt with the main problems in the pure-hydrogen fuel cell system for the vehicular application. It is valuable for the design, system integration and control of the proton exchange membrane fuel cell systems.
