基于POS算法优化的PID神经网络的系统控设计.doc

中北大学
研究生《神经网络及应用》作业

课题名称
中文基于PSO算法优化的PID神经网络的系统控设计
英文Research on the Design of Algorithm Optimization of PSO of work PID Control system


姓名王强龙
学号 S20100622
班级 Y100202
专业模式识别与智能系统
研究方向基于网络的智能控制
所在院、系机械工程及其自动化学院
基于PSO算法优化的PID神经网络的系统控设计
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。本文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,文中系统的介绍了两种种改进方式,主要有:遗传算法PID控制器和神经网络PID控制器。
神经网络具有强的非线性映射能力、自学忆能力、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。文中主要对基于单神经元PID控制器、BP神经网络PID控制器进行研究。对于BP神经网络初始权值选择困难的问题,本文采用粒子群优化算法(PSO)来对BP神经网络控制器进行优化。本文同时也利用PSO算法对常规PID控制器的参数进行整定研究。
最后,本文对单神经元自适应PID控制系统和基于PSO优化的BP神经网络PID控制系统进行仿真试验,发现后者使系统的性能有所提高。
关键词:神经网络PID控制器;BP算法;PSO算法;
Research on the Design of Algorithm Optimization of PSO of work PID Control system
Abstract
The technique of PID control is very general, and it is applied in many fields at present. In the paper, work theory foundation, studying algorithm of the work and traditional PID controller are introduced, and traditional PID controller weak point is analyzed in controlling plicated, dynamic and uncertain system. In order to achieve the goal of improving traditional PID controller, two kind improvement ways are put forward in the paper— hereditary algorithm PID controller and work PID controller.
Because of the strong nonlinearity to shine upon ability, study adaptive capacity, associative memory ability, processing method of proceed information and fine fault-tolerant performance, there are better control results to plicated system in industrial control, and the unstable control results caused by the change of system structure or parameters have been improved after the PID controller improved by
work. Single work PID controller and BP work PID controller have been study chiefly in the paper. Because there are some difficult in choosing the first power value of the BP work, PSO algo