实用随机多变量NARMAX模型的无模型自校正控制器  

Self-tuning controller with model-free of practical random multivariable NARMAX model

作　　者：侯小秋 李丽华 HOU Xiaoqiu;LI Lihua(School of Electronics and Controlling Engineering,Heilongjiang University of Science and Technology,Haerbin 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）》2023年第5期8-13,共6页Journal of Shaanxi University of Technology:Natural Science Edition

摘　　要：针对实用随机多变量NARMAX模型的控制问题,基于改进多变量广义控制目标函数,提出了实用随机多变量NARMAX模型的无模型自校正控制器。使用具有辅助向量的多变量全格式动态线性化方法逼近实用随机多变量NARMAX模型,由其构建多变量预测模型,采用BP神经网络辨识泛模型参数函数,增加了泛模型参数函数的信息含量,将多变量系统分离为具有耦合的多个子系统,通过非线性递推最小二乘法对BP神经网络的连接权重值进行学习,同时估计随机干扰模型的参数,给出了新的学习算法。仿真研究表明算法的响应性能优良。For the control problem of practical random multivariable NARMA model,based on modified multivariable universal control function,a model-free self-tuning controller of practical random multivariable NARMA model was developed.Multivariable full form dynamic lineation with auxiliary variable was applied to approach the practical random multivariable NARMA mode,and its multivariable predictive model was built.BP neural network was used to identify universal model parameter function and the information content of universal model parameter function was increased.Multivariable system was separated into multi-subsystem with couplingand connect weight value of BP neural network identification was learned by nonlinearity recursive least squares method,also random interference model parameters were estimated at mean time,a new learning algorithm was obtained.Simulation results indicate that the system response have excellent performance.

关 键 词：神经网络控制 无模型自适应控制 自校正控制器 非线性递推最小二乘法 参数估计 

分 类 号：TP273.2[自动化与计算机技术—检测技术与自动化装置]