基于神经网络非线性模型的多级工作点阶跃响应扩展DMC预测控制  

Extension of DMC Predictive Control Using Neural Network Based Nonlinear Models andMulti-step Operation Point Respones

作　　者：刘军[1] 何星[1] 许晓鸣[1] 

机构地区：[1]上海交通大学自动化研究所,200030

出　　处：《控制与决策》2000年第3期342-344,共3页Control and Decision

摘　　要：利用前馈神经网络建立对象的非线性预测模型 ,在不同工作点做阶跃响应 ,建立其局部线性模型。用隶属函数将局部线性模型加权得到全局线性模型。全局线性模型用于滚动优化 ,非线性模型用于预测系统输出和校正线性模型 ,实现非线性预测控制。仿真结果表明该方法控制效果良好 ,可满足实时要求。Feedforward neural networks for modeling a nonlinear system are used to obtain its nonlinear model. Local linear models of a nonlinear system are built using multi-step respones at several operation points, and global linear model equals to weight local linear models by membership functions. Nonlinear predictive control is realized by the global linear model based roll optimizing, and ontime adjusting using neural network based nonlinear model of the nonlinear system. Simulation results demonstrate that the proposed control system has good performance and meets real time demand.

关 键 词：神经网络 多级工作点 非线性模型 DMC预测控制 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置] O231[自动化与计算机技术—控制科学与工程]