基于分数阶优化的神经网络自适应解耦控制算法  

Neural Network Adaptive Decoupling Control Algorithm Based on Fractional Order Optimization

作　　者：张继研[1] 王宏伟[1] 徐占国[1] ZHANG Jiyan;WANG Hongwei;XU Zhanguo(Experimental Teaching Center of Information Technology,Dalian University of Technology,Dalian Liaoning 116024,China)

机构地区：[1]大连理工大学信息技术实验教学中心,辽宁大连116024

出　　处：《信息与电脑》2021年第16期81-84,共4页Information & Computer

基　　金：大连理工大学本科教学改革项目“基于自制设备的‘全开放PLC口袋实验室建设’”(项目编号:YB2021008)。

摘　　要：针对多变量、强耦合、杂非线性的复杂系统,笔者提出了一种基于分数阶优化的神经网络自适应解耦控制算法。首先,提出了基于分数阶梯度下降算法来优化RBF神经网络模型参数,并将其应用到RBF神经网络解耦控制系统中;其次,通过数值仿真实例和浸没式电极锅炉实例来证明提出算法的有效性。仿真结果表明分数阶算法可以提高RBF神经网络解耦控制器的收敛性,具有响应速度快、调节时间短以及超调量小等特点。A neural network adaptive decoupling control algorithm based on fractional order optimization is proposed for multivariable,strongly coupled and nonlinear complex systems.Firstly,a fractional step descent algorithm is proposed to optimize the parameters of the RBF neural network model and applied to RBF neural network decoupling control system;Secondly,the effectiveness of the proposed algorithm is proved by numerical simulation and submerged electrode boiler.The simulation results show that the fractional order algorithm can improve the convergence of the RBF neural network decoupling controller,and has the characteristics of fast response,short adjustment time and small overshoot.

关 键 词：分数阶算法 RBF神经网络 解耦控制 PID控制 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]