神经网络辨识的无模型自适应自校正控制器  

Adaptive self-tuning controller with model free of neural net identification

作　　者：侯小秋[1] 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3期9-14,76,共7页Journal of Shaanxi University of Technology:Natural Science Edition

摘　　要：在针对全格式动态线性化泛模型中引入辅助变量,提出了一种改进的全格式动态线性化泛模型,克服了其存在的问题。采用BP神经网络对其进行辨识,基于广义目标函数,提出神经网络辨识的无模型自适应隐式自校正控制器,其算法是关于当前控制输入的非线性方程,利用牛顿-拉夫逊算法求解,根据直接极小化指标函数的自适应优化算法对BP神经网络的连接权重值进行在线学习。仿真研究验证了所提出的隐式自校正控制器的有效性,系统具有良好的控制品质。By introducing auxiliary variable the dynamic linearization universal model,a modified model has been developed,which overcomes the problems existed in original model.Its identification is carried out by using BP neural network.Based on generalized objective function,a model-free adaptive implicit self-tuning controller of BP neural network has been derived.And its algorithm is regarded to nonlinear equation of current control input,which is solved by Newton-Raphson method.Connection weight value of BP neural network is online learned by using the adaptive optimization algorithm for direct minimization of index function.Simulation results verify the validation of the implicit self-tuning controller and the system having excellent control quality.

关 键 词：神经网络控制 无模型自适应控制 自校正控制 非线性系统 牛顿-拉夫逊算法 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]