一种改进RBF网络与PID相协调的自适应控制  

Self-adaptive control with coordinated PID and improved RBF network

作　　者：许昌[1] 吕剑虹[1] 程明[1] 郑源[2] 

机构地区：[1]东南大学电气学院,江苏南京210096 [2]河海大学水电学院,江苏南京210098

出　　处：《哈尔滨工程大学学报》2007年第6期660-664,共5页Journal of Harbin Engineering University

基　　金：教育部博士点基金资助项目(20050286041);河海大学科技创新基金资助项目(2006406086)

摘　　要：为充分利用人工神经网络处理非线性问题能力和PID控制器设计简单以及鲁棒性强的优点,设计了高性能控制器.通过对最小资源分配网络的裁剪策略进行改进,使改进的网络具有更加紧凑的结构;把改进的最小资源分配网络应用到神经网络自适应控制器中,同时结合神经网络自适应控制和PID控制,提出一种神经网络与PID相协调的自适应控制策略,并应用到对电站锅炉过热汽温的控制上.仿真结果表明,提出的控制算法与常规的控制算法相比具有更好的控制品质..A high performance controller is devised to make full use of neural network capacity in dealing with nonlinear problems by applying a PID controller＇s advantages of simple design and strong robustness. A PID neural network （PIDNN） control system consists of three hidden layers of proportional （P）, integral （I）, and derivative （D） neurons. A PIDNN＇s weights are adjusted by back-propagation algorithms. Through an improved pruning strategy for a minimal resource allocating network （MRAN）, the resulting network is made more compact. The improved MRAN is applied to the neural adaptive control （NAC） strategy. Combining NAC with PID control, a self-adaptive control strategy is presented which coordinates the neural network and PID. This strategy was used to simulate control of the superheated steam temperature of a power station boiler. Results show the algorithm performs better than conventional control algorithms.

关 键 词：最小资源分配网络 神经自适应控制 过热汽温 

分 类 号：TP11[自动化与计算机技术—控制理论与控制工程]