【作者】 王艳秋；

【导师】 顾树生； 王建辉；

【作者基本信息】 东北大学 ， 控制理论与控制工程， 2006， 博士

【摘要】 迄今为止，已在工业上获得成功应用的先进控制策略主要包括：自适应控制、变结构控制、预测控制、鲁棒控制、模糊控制、专家控制、神经网络控制等，各种控制策略各有利弊。因此多种控制策略之间相互渗透、交叉和结合的复合控制策略，是先进控制技术的发展方向。这些复合控制策略克服了单一控制策略的不足，取长补短、具有更优良的性能，能更好地满足不同应用场合的不同要求，因而获得了更广泛的应用。 所谓先进控制技术的集成包括两个方面：一方面是将几种先进的智能控制方法融合在一起，构成具有高度自主能力的智能混合控制系统；另一方面是将智能控制与最优控制相结合，构成智能复合控制，以便取长补短，优势互补，提高整体控制水平。因此智能混合控制和智能复合控制在非线性多变量控制系统中的应用已经成为当前控制领域的一个研究热点，并已成为解决各类复杂系统控制问题的重要工具，在理论上和实际应用中都有十分重要的意义。 本文在掌握国内外研究现状的基础上，在这一领域展开了进一步的研究工作，主要研究内容概括如下： (1)描述了模糊系统与神经网络的等价性和互换性：从映射角度来看，模糊系统和神经网络都具有非线性函数的逼近能力；模糊系统与神经网络具有等价性和互换性：模糊系统可以用一等价的神经网络来表示，同样神经网络也可以用一等价的模糊系统来表示；其次，描述了模糊系统和神经网络融合的五种形态，并提出一种结构等价的模糊神经网络，给出了它们的实现方法。 (2)研究了智能混合控制。并以改进的遗传算法在模糊神经网络参数学习中的应用问题为例加以说明。首先提出了两种改进的遗传算法：一种是加入BP算子的遗传算法，这不仅利用了GA与BP各自的优点，而且极大的提高了参数的搜索效率；另一种是改变遗传算子的遗传算法，并证明了这种改进的遗传算法的性能明显优于标准的遗传算法。然后将改进的遗传算法应用于模糊神经网络参数学习中，通过构造模糊神经网络控制器进行仿真，验证了方法的有效性和适用性。改进的遗传算法在学习过程中表现出很强的全局优化能力，为复杂非线性及组合优化问题提供了一个很好的解更多还原

【Abstract】 So far, the advanced control technology strategy has obtained successful application in the industry and mainly includes: adaptive control, variable structure controls, predictive control, robust control, fuzzy control, expert control, neural network control and so on. Each kind of control strategy has its special skill respectively and has certain problems in certain aspects. Therefore, between each kind of control strategy mutually seeping, overlapping and the union. Moreover the new compound control strategy is continuously emerging along with the science and technology development. These compound control strategies overcome the insufficiency of single control strategy, making up for one’s deficiency by learning from others’ strong points, and have a finer performance. These strategies can even better satisfy the different application situation’s request and obtain a more widespread application.Integration of advanced control includes two aspects. One is the fusion of several advanced intelligent control methods. It constructs advanced hybrid intelligent control system, which has highly independent capacity. The other is the integration of intelligent control technology and traditional control theory. It constructs intelligent compound control system, and has the advantage of improving the whole control level. So the application of advanced hybrid intelligent control and intelligent compound control in nonlinear multivariable control system has become a hotspot in current control field and an important tool for resolving all kinds of complex system control problems. It has very important meaning in theory and actual application.This paper is on the basis of the domestic and foreign research status, further the research work in this domain. And the main contents are as follows.(1) To describe equivalence and interchangeability of the fuzzy system and neural network : Firstly, the fuzzy system and the neural network all have approximation ability of更多还原