基于Elman网络预测的神经网络控制器在冷轧APC系统中的应用研究  

Neural Network Controller for APC System Constructed Based on Elman Network Prediction

作　　者：魏立新[1] 林鹏[1] 王利平[1] 

机构地区：[1]燕山大学工业计算机控制河北省重点实验室,河北秦皇岛066004

出　　处：《矿冶工程》2015年第6期130-134,139,共6页Mining and Metallurgical Engineering

基　　金：国家自然科学基金钢铁联合基金资助项目(U1260203)

摘　　要：为解决冷轧自动位置控制(APC)系统的非线性控制问题,提出了一种基于Elman网络动态辨识的神经网络控制算法。利用改进的Elman网络建立APC系统的辨识模型,对系统的非线性特性进行动态辨识;采用一个前馈神经网络输出控制量,并根据Elman网络的辨识结果相应调整控制策略;借鉴预测控制中柔化设定值和反馈校正的思想提高系统响应性能。MATLAB对比实验表明:该方法控制速度快、响应精度高。To solve nonlinear control problems in the automation position control（ APC） system for cold rolling process,a network control algorithm based on Elman neural network identification was proposed. With an improved Elman network,an identification model for APC system was constructed to identify dynamically the nonlinear characteristics of the system. The control strategy can be correspondingly adjusted according to the control signal outputted by a feed-forward neural network and identification result of Elman network. The response performance of system can be improved by referring to some ideologies in predictive control such as setpoint softening and feedback correction. MATLAB comparative tests testified the fast response and high accuracy of the proposed method.

关 键 词：自动位置控制 冷轧 ELMAN神经网络 辨识 神经网络控制 预测控制 

分 类 号：TG334[金属学及工艺—金属压力加工]