【作者】 赵承滨；

【导师】 宋清昆；

【作者基本信息】 哈尔滨理工大学 ， 控制理论与控制工程， 2004， 硕士

【摘要】 遗传算法、神经网络与模糊控制是当前人工智能中的主要研究领域。由于它们在特性上有许多共同性和互补性，将遗传算法、神经网络与模糊逻辑相结合的研究已成为当前的研究热点之一。本文旨在探讨将遗传算法与神经网络、模糊控制相结合以获得智能信息处理功能的途径，并探讨了基于遗传算法的模糊神经控制在工业锅炉燃烧系统中的应用。本文以“遗传算法→模糊神经网络→基于遗传算法的模糊神经网络”为主线，全文共分三部分：遗传算法、模糊神经网络和基于遗传算法的模糊神经网络。主要内容如下：第一、遗传算法的基本理论首先通过对常见的搜索方法及其特点的比较，阐述了遗传算法的特点及研究遗传算法的必要性。然后介绍了遗传算法的产生和发展、主要研究成果和今后研究的主要课题等。着重阐述了遗传算法的基本原理、流程及GA的MATLAB实现问题。最后以工业燃烧系统为例，讨论了遗传算法的应用问题。通过对基于遗传算法的工业锅炉燃烧系统热效率自寻优控制系统的研究表明：作为一种优化算法，遗传算法具有原理简单、易于实现、寻优速度快、鲁棒性强等特点。第二、神经网络与模糊控制相结合的模糊神经控制理论针对模糊神经网络理论，以“生物解剖学的神经元→控制领域的神经元模型→神经网络→模糊神经网络”为主线：先介绍了传统的神经网络的结构、功能、基本属性及学习规则等；再介绍模糊神经网络的拓扑结构、分类、隶属函数的特点及相应激励函数的确定方法等内容；最后，以模糊小脑神经网络(FCMAC)为算例阐述了模糊神经网络的应用问题，函数仿真结果表明：与单独采用神经网络相比，采用模糊神经网络具有更高的精确度和更快的学习速度。因此，它具有更广泛的使用和研究价值。第三、基于遗传算法的模糊神经控制理论这部分主要探讨将遗传算法与模糊控制、神经网络相结合的控制策略，首先是将遗传算法用于模糊控制，主要表现在模糊遗传算法和遗传算法用于模糊优化两方面。其次是遗传算法在神经网络方面的应用：主要是在网络学习、网哈尔滨理_厂大学卜学硕士论文络设计和网络分析方面的应用。最后介绍了遗传算法和模糊神经网络相结合的控制策略，并以锅炉燃烧系统为算例，讨论了基于遗传算法的模糊神经控制的应用问题，仿真结果表明，该法具有寻优速度快、鲁棒性强等优点。 遗传算法、神经网络与模糊控制相结合的控制理论具有显著的优点，表明它在智能控制领域中有着巨大的应用潜力，因此加速人工智能领域的研究是一个非常重要的研究方向。更多还原

【Abstract】 Genetic Algorithms, Neural Network and Fuzzy Control are important research fields in artificial intelligence. The research on combinations of genetic algorithms, neural network and fuzzy logic is attracting the attention of many researchers because of many common and complementary features among them. The purpose of this thesis is to discuss the ways to combine genetic algorithms, neural network and fuzzy logic control and explore how to acquire the channel of intelligent information processing. Moreover, the application of the fuzzy neural network based on genetic algorithms in the field of industry boiler combusting system is discussed.The clue of this paper is genetic algorithms fuzzy neural network fuzzy neural network based on genetic algorithms. The whole paper concludes three parts: genetic algorithms, fuzzy neural network and fuzzy neural network based on genetic algorithms. The content of this thesis is as follows:1. The basic theory of genetic algorithmsFirstly, through comparing the familiar researching methods and characteristics, the characteristics of GA and the necessary of researching GA are discussed. Then, the generation & progress, the prime researching fruits, the important researching tasks in future in the field of GA are discussed. The significant content in this part lies in the basic principle, flow and the realization in MATLAB of GA. Finally, as the example of industry boiler combusting system; the application of GA is discussed. Through the researching of the industry boiler combusting thermal efficiency self-optimized system , we can draw a conclusion: as a kind of self-optimized method, generic algorithms has the virtues of the simple principle, the easy realization, prompt self-optimized paces and robustness.2. The fuzzy neural control theory combining neural network and fuzzy controlFor the fuzzy neural control theory, the clue of this part is the neuron in anatomy-the neuron model in the field of control neural network fuzzy neural network. First, the structure, function, basic attribution and learning rules of neural network are discussed. Next, the topological structure, classification, the characteristic of membership function and the definition methods of excitation function in the field of neural network are discussed. Finally, as the example of Fuzzy Cerebella Model Articulation Controller, the application of neural network is discussed. The result of simulink indicated that comparing with Cerebella Model Articulation Controller, the FCAMC have the virtues of higher precision and the faster-paces. As a result, it has the more widely researching value. 3. The fuzzy neural control theory based on genetic algorithms The main ideas of this part lie in the control strategy that combine GA & fuzzy control and combine GA & neural network. Firstly, genetic algorithms are applied into fuzzy control. It lies in the fuzzy genetic algorithms and fuzzy optimizing using GA. Then, genetic algorithms are combined into neural network. It lies in the three aspects as follows: network learning, network designation and network analysis. Finally, the control strategy, which combines GA and fuzzy neural network, is discussed. In addition, as an example of the boiler combusting system, the application of fuzzy neural control based on genetic algorithms is discussed. The result of simulink indicated that it has the virtue of the faster-paces and robustness.The control theory of combining genetic algorithms into fuzzy control and neural network has the remarkable virtue. It indicates the immense application potential in the field of intelligent control. As a result, it is accelerating the research in the field of artificial intelligent that is a significant research direction in future.更多还原