【作者】 曾军；

【导师】 方厚辉；

【作者基本信息】 湖南大学 ， 电工理论与新技术， 2004， 硕士

【摘要】 PID控制技术是一种应用很普遍的控制技术，目前在很多方面都有广泛的应用。论文深入研究了神经网络PID控制器。首先简要介绍了神经网络的理论基础和神经网络的学习算法，传统的常规PID控制器，针对常规PID控制器对于复杂的、动态的和不确定的系统控制还存在着许多不足之处进行了分析，为了达到改善常规PID控制器在复杂的、动态的和不确定的系统控制还存在着许多不足之处的目的，文中系统的提出了五种改进方式(模糊PID控制器、专家PID控制器、灰色PID控制器、遗传算法PID控制器和神经网络PID控制器)。 本文主要研究了神经网络PID控制器。利用神经网络具有强的非线性映射能力、自学习能力、联想记忆能力、并行信息处理方式及优良的容错性能，应用神经网络对PID控制器进行改进后，对于工业控制中的复杂系统控制有着更好的控制效果，有效的改善了由于系统结构和参数变化导致的控制效果不稳定。文中深入研究了基于神经网络的神经元PID控制器、BP算法神经网络PID控制器，并针对BP算法在应用时易陷入局部极小点，而且学习时间长甚至达不到学习的目的提出了应用模糊算法、遗传算法对BP网络进行改进的组合优化神经网络PID控制器。最后，对常规PID控制器和神经网络PID控制器进行了仿真比较，仿真结果表明，应用神经网络对常规PID控制器进行改进后提高了系统的鲁棒性和动态特性，有效的改善了系统的控制结果，达到了预期的目的。更多还原

【Abstract】 The techologic of PID control is very general, there is extensive application in many fields at present. The paper further investigates the neural network PID controller. Have introduced the theoretical foundation of the neural network and studying algorithm of the neural network, traditional routine PID controller briefly at first, is it analyse to traditional PID controller to complicated, dynamic and uncertain system control for a lot of weak points to it to go on, in order to achieve the goal of improving traditional PID controller, five kinds of improvement ways of systematic proposition in the article (fuzzy PID controller, expert PID controller, grey PID controller, hereditary algorithm PID controller and neural network PID controller).The text has discussed the neural network PID controller mainly. Because the neural network has strong nonlinearity to shine upon ability, study adaptive capacity, associative memory ability, the information processing way run side by side by oneself and fine fault-tolerant performance, there are better control result to the complicated system in industrial control, improveing the structure of system effective and change of the parameter result in controlling the unstability of the result. The paper further investigate the neuron PID controller based on neural network, neural network PID controller of BP algorithm, and algorithm while using easy to put some snack very much into to BP, but also has learnt the chronic purpose to even reach study and proposed using the fuzzy algorithm, the hereditary algorithm carries on the improved association and optimizes the neural network PID controller to BP network. Finally, carried on simulation to traditional PID controller and neural network PID controller, simulation result indicate, use neural network go on after improving raising stupid getting wonderful and dynamic characteristicking of system to routine PID controller, the effective systematic control result of improvement, has achieved the anticipated goal.更多还原