RBF神经网络整定参数的预测滤波PID控制  

Predictive filtering PID control of radial basis function neural network parameter setting

作　　者：侯小秋 HOU Xiao-qiu(School of Electronics and Controlling Engineering,Heilongjiang University of Science and Technology,Harbin 150022,China)

机构地区：[1]黑龙江科技大学电气与控制工程学院,黑龙江哈尔滨150022

出　　处：《陕西理工大学学报（自然科学版）》2022年第1期33-39,共7页Journal of Shaanxi University of Technology:Natural Science Edition

摘　　要：针对非线性系统采用常规PID控制效果不佳的问题,通过径向基函数(RBF)神经网络构建系统的辨识器,推导了辨识器输出关于RBF神经网络参数的梯度及二阶导数矩阵。采用直接极小化指标函数自适应算法,实现RBF神经网络的输出权和节点中心及节点基宽参数的在线学习算法。基于增量式预测滤波PID控制,应用最速下降法对增量式预测滤波PID控制参数进行在线优化,提出基于RBF神经网络在线整定参数的预测滤波PID控制算法。仿真研究表明,因算法具有在线整定参数和预测控制性能,故具有良好的动、静态性能。In the light of poor performance of nonlinear system employing conventional PID control,using a RBF radial basis function neural network structures system identifier,the gradient and second-order derivative matrix of the radial basis function(RBF)neural network parameter were deduced.Adopting direct minimization index function adaptive algorithm,an online learning algorithm of output connection weight,node center and node basis width parameter of RBF neural network were obtained.Based on incremental predictive filtering PID control,the study used gradient descent method to optimize online control parameter of incremental predictive filtering PID.A predictive filtering PID control algorithm based RBF radial basis function of RBF neural network parameter setting was developed.Simulation results indicate that the algorithm exhibits the dynamical state and static state characteristics due to the functions of online parameter setting and predictive control.

关 键 词：径向基函数神经网络 非线性系统 预测滤波PID控制 PID控制参数整定 在线学习算法 直接极小化指标函数自适应算法 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]