【作者】 李信洪；

【导师】 王劭伯；

【作者基本信息】 福州大学 ， 控制理论与控制工程， 2002， 硕士

【摘要】 本文从工程实际出发，对基于神经网络的时滞控制系统设计理论和应用进行了研究。文中着重研究了神经元网络设计理论、基于神经网络的线性和非线性时滞对象辨识及时滞系统控制方法，以此为基础提出了时滞系统神经元自适应PID内模控制、时滞系统神经网络自适应PIP控制以及神经网络自适应Smith控制方法，并进行了控制系统的实例设计与仿真实验。 本文的主要研究内容有如下几个方面： 1．综述了神经网络、神经控制及时滞系统的研究现状，并就基于神经网络的时滞系统控制方法提出了作者的一些观点； 2．介绍了时滞系统控制的传统方法，包括PID控制及其改进算法、Smith预估控制、PIP控制等，并通过仿真实验说明了传统时滞系统控制方法的弱鲁棒性； 3．探讨了面向控制的神经元网络设计理论，包括单神经元控制的结构和基本理论及BP神经网络设计； 4．针对时滞系统的辨识问题，介绍了传统的时滞辨识方法，提出了两种基于神经网络的辨识方法，包括时滞线性系统的辨识和时滞非线性系统的辨识，并对神经网络辨识与传统辨识方法进行了比较； 5．将内模型控制与Smith预估控制相结合，并引入单神经元控制器，提出了时滞系统神经元自适应PID内模控制方法，并将其应用于电加热炉的温度控制，取得了较好的控制效果； 6．探讨了神经网络自适应内模控制与神经网络预测PID控制方法，并分别进行了控制系统的设计与仿真实验； 7．基于神经网络辨识方法，提出了两种神经网络时滞系统补偿控制策略，即神经网络自适应PIP控制与神经网络自适应Smith补偿控制，并进行了控制系统的设计与仿真实验，取得了很好的控制效果。更多还原

【Abstract】 From the points of view for practical uses, the theory and applications of time-delay control systems based on neural network are researched in this dissertation. The design theory of neural networks and the estimations of linear systems and nonlinear systems with dead time using neural network are discussed. Neuro-PID internal model control, adaptive PIP controller based on neural network and adaptive Smith Predictor using neural network are proposed for time-delay systems. The simulation experiments and the real-time temperature control of electrical furnace are made.The main research work and contributions of this dissertation are as follows:1. A survey of neural networks, neuro-control and time-delay systems is summarized, and the problems existing in time-delay control systems are also discussed.2. An introduction of the traditional design methods for time-delay systems is given, including a PID controller, a Smith Predictor and a PIP controller. And the pool robustness of these traditional design methods is pointed out by the simulation experiments.3. The design theory of neural networks is discussed, including the basis principles of neuron control and the design of Back-Propagation Network.4. To the identification of time-delay systems, the traditional identification methods are introduced. Two kinds of identification methods are put forward, including linear systems and nonlinear systems with dead time. The comparison between the traditional identification methods and the ways based on neural network is made.5. Combining internal model control method with Smith Predictor and single neuron controller, a neuro-PID internal model control method for time-delay systems is put forward. The real-time temperature control result of the electrical furnace shows that the control performance is very good.6. Adaptive internal model control and predictive PID control method are discussed, and the simulation experiments are made.7. According to the identification methods based on neural network, two kinds of compensative control strategies for time-delay systems are proposed, namely adaptive PIP controller based on neural network and adaptive Smith Predictor using neural network, and the simulation experiments show that the control performance is very good.更多还原