基于分数阶的神经网络解耦控制优化方法  

Optimization Method for Neural Network Decoupling Control Based on Fractional Order

作　　者：宋帆 马小晶[1] 王宏伟[2] 陈洁[1] 贺航 SONG Fan;MA Xiao-jing;WANG Hong-wei;CHEN Jie;HE Hang(School of Electrical Engineering,Xinjiang University,Urumqi 830047,China;School of Control Science and Engineering,Dalian University of Technology,Dalian 116024,China)

机构地区：[1]新疆大学电气工程学院,新疆乌鲁木齐830047 [2]大连理工大学控制科学与工程学院,辽宁大连116024

出　　处：《控制工程》2022年第4期692-698,共7页Control Engineering of China

基　　金：国家自然科学基金资助项目(12002296,61863034)。

摘　　要：针对强耦合、多变量的非线性系统,提出了一种基于Caputo分数阶微分优化的BP-PID解耦控制算法。首先,应用Caputo定义的分数阶思想设计分数阶梯度下降算法,并将其应用到BP-PID控制系统,以实现多变量耦合系统的解耦控制;其次,通过测试的二维变量函数验证所提算法的收敛性;最后,在浸没式电极锅炉耦合模型中使用分数阶梯度下降算法优化的BP-PID算法,并与基于传统梯度下降算法的BP-PID算法进行对比。实验结果表明,所提算法提高了BP-PID解耦控制器的收敛速度,并且加快了响应速度,减少了超调量,缩短了调节时间。Absrtact: Aiming at nonlinear systems with strong coupling and multiple variables, a algorithm of BP-PID decoupling control algorithm based on Caputo fractional order differential optimization is proposed in this paper. Firstly, the fractional order gradient descent algorithm is designed by applying the fractional order idea defined by Caputo, and applied to a BP-PID decoupling control system to realize the decoupling control of the multi-variable coupled system. Secondly, the convergence of the proposed algorithm is verified by the tested two-dimensional variable function. Finally, the BP-PID algorithm optimized by the fractional gradient descent algorithm is used in the coupled model of the submerged electrode boiler, and compared with the BP-PID algorithm based on the traditional gradient descent algorithm. The experimental results show that the proposed algorithm improves the convergence speed of the BP-PID decoupling controller, accelerates the response speed, reduces the overshoot, and shortens the adjustment time.

关 键 词：分数阶 梯度下降 BP神经网络 解耦控制 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]