强非线性系统的一种快速神经网络控制策略  

A FAST NEURAL NETWORK CONTROL STRATEGY OF A SEVERE NONLINEAR SYSTEM

作　　者：张海涛[1] 陈宗海[1] 向微[1] 秦廷[1] 王雷[1] 

机构地区：[1]中国科学技术大学自动化系,合肥230027

出　　处：《模式识别与人工智能》2003年第4期385-389,共5页Pattern Recognition and Artificial Intelligence

基　　金：2002年度国家985高水平大学建设基金;中国科学技术大学青年基金(No.KB1025);2002年度合肥市重点科技计划等资助

摘　　要：采用传统的神经网络逆模策略控制具有强非线性的系统,因其计算量过大导致在线实时性能不佳,本文提出一种新型快速径向基神经网络在线逆模控制策略,并利用锥度准则对控制系统的稳定性进行了理论分析,对强非线性对象的控制仿真结果表明,在保证控制精度的前提下,该算法大大提高了控制器运算的速度,且对扰动、时延、非线性及对象参数的摄动有较强的适应能力,具有较好的控制品质,适合应用于复杂工业过程控制器的设计.The realtime performance is not satisfying when traditional neural network backward model control strategy is applied to severe nonlinear system, because the calculated amount of this strategy is too large. A novel online backward model control strategy based on a fast radial basis neural network is brought forward in this paper. The stability analysis of this control system using Conicity Criterion is presented as well. The simulation results on a severe nonlinear plant show that this strategy can greatly increase the controller's calculating speed while the control precision index is ensured. Relative merits such as good adaptation to disturbance, time delay, nonlinear and the drift of plant parameters are given. So excellent control performance can be gained from this strategy, and the strategy is suitable for applying to the design of complicated industrial process controller.

关 键 词：强非线性系统 神经网络 控制策略 逆模控制 控制精度 非线性系统 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]