引用本文:张宇,程开新,竺俊杰,武国勋,姚熊亮.级联RTAC系统动态神经网络辨识与分散镇定控制[J].控制理论与应用,2022,39(8):1451~1459.[点击复制]
ZHANG Yu,CHENG Kai-xin,ZHU Jun-jie,Wu Guo-xun,YAO Xiong-liang.Dynamic neural network identification and decentralized stabilization control of cascade RTAC system[J].Control Theory and Technology,2022,39(8):1451~1459.[点击复制]
级联RTAC系统动态神经网络辨识与分散镇定控制
Dynamic neural network identification and decentralized stabilization control of cascade RTAC system
摘要点击 3080  全文点击 337  投稿时间：2021-07-08  修订日期：2022-09-08
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2022.10602
  2022,39(8):1451-1459
中文关键词  级联RTAC  动态神经网络  分散控制  不确定关联项  辨识
英文关键词  cascade RTAC  dynamic neural network  decentralized control  uncertain interconnected term  identification
基金项目  国家自然科学基金项目(5197090325, 51809054), 黑龙江省自然科学基金项目(LH2020E075)资助.
作者单位E-mail
张宇 哈尔滨工程大学船舶工程学院 hblwzy@hrbeu.edu.cn 
程开新 大连理工大学电气工程学院  
竺俊杰 哈尔滨工程大学船舶工程学院  
武国勋* 哈尔滨工程大学船舶工程学院 wuguoxun@163.com 
姚熊亮 哈尔滨工程大学船舶工程学院  
中文摘要
      针对含不确定关联项的级联RTAC系统的镇定控制问题, 提出了一种基于动态神经网络辨识的分散控制方 案. 应用拉格朗日方程建立起了考虑不确定非线性作用力的级联RTAC系统数学模型, 采用动态神经网络实现级 联RTAC系统中不确定关联项的在线辨识, 通过构造含神经网络权值矩阵迹的Lyapunov函数, 证明了辨识误差的一 致有界性. 通过动态神经网络辨识不确定关联项、补偿系统建模误差, 建立级联RTAC系统分层滑模控制算法, 以实 现级联RTAC系统的高精度分散镇定控制. 数值仿真验证了动态神经网络的引入对级联RTAC系统分散镇定控制系 统瞬态幅值抑制、稳态精度提升的效果.
英文摘要
      A dynamic neural network based decentralized control scheme is proposed for the stabilization of cascade RTAC system. The mathematical model of the cascade RTAC with uncertain interconnected forces is derived. The dynamic neural network is adopted for identification of uncertain interconnected terms in the mathematical model, and uniform boundedness theorem of identification errors is proved, via introducing Lyapunov function including the trace of weight matrix of dynamic neural network. Then, a decentralized stabilization control scheme of cascade RTAC system is designed using the hierarchical sliding mode algorithm to precisely stabilize the cascade RTAC system, based on dynamic neural network identification and compensation of modeling error. Numerical simulations are conducted to prove the effectiveness of the proposed control scheme with the introduction of dynamic neural network identification in suppression of vibration amplitude and control precision.