一种神经网络PID自适应控制及其应用研究  

Neural Network PID Adaptive Control and Its Application

作　　者：许昌[1] 吕剑虹[1] 程明[1] 郑源[2] 

机构地区：[1]东南大学电气工程系,江苏南京210096 [2]河海大学热能与动力工程系,江苏南京210098

出　　处：《控制工程》2007年第3期284-286,共3页Control Engineering of China

基　　金：国家教育部博士点基金资助项目(20050286041)

摘　　要：为提高PID控制的自适应性能,提出了一种神经网络PID自适应控制算法。最小资源分配网络(MRAN)是一种可以在线调整隐节点的RBF网络,具有泛化能力强、计算精度高的优点。通过对MRAN的剪裁策略进行改进,使改进的网络具有更加紧凑的结构;把改进的MRAN和PID控制相结合,提出了一种基于动态RBF网络的PID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,所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 neural network PID control algorithm is presented to improve the PID adaptive control performance. Minimal resource allocating network （MRAN）that is an RBF network of self-tuning hidden nodes online and has the excellences of strong generation and high accuracy. The structure is improved by the reducing strategies of MRAN. The improved MRAN has more compact structure. Combining the improved MRAN with PID control, the PID self-adaptive control on dynamic RBF network is presented, and applied to control the superheated steam temperature of a power station boiler. Simulation results show that the presented algorithm has better performances than the conventional control algorithm.

关 键 词：最小资源分配网络 PID自适应控制 过热汽温 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]