基于前馈神经网络的分数阶事件触发控制器实现及应用研究  

Implementation and application of fractional-order event-triggeredcontroller based on feedforward neural network

作　　者：于南翔[1,2] 朱伟[2] YU Nanxiang;ZHU Wei(School of Computer Science and Technology,Chongqing University of Posts and Telecommunications,Chongqing 400065,P.R.China;Key Laboratory of Intelligent Analysis and Decision on Complex Systems,Chongqing University of Posts and Telecommunications,Chongqing 400065,P.R.China)

机构地区：[1]重庆邮电大学计算机科学与技术学院,重庆400065 [2]重庆邮电大学复杂系统智能分析与决策重点实验室,重庆400065

出　　处：《重庆邮电大学学报（自然科学版）》2023年第6期1072-1079,共8页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金：国家自然科学基金项目(61876200,62373071);重庆市教委科学技术研究计划项目(KJZD-K202000601,KJZD-M202100602);重庆市自然科学基金项目(cstc2019jcyj-msxmX0102);重庆邮电大学自然科学基金(A2008-50);重庆邮电大学博士生培养项目(BYJS201914)。

摘　　要：为了提高事件触发控制算法的普适性,降低现有控制算法对具体数学表达式的依赖,提出了一种基于前馈神经网络的事件触发控制方法,实现了更具一般化的分数阶事件触发控制器设计。该方法充分发挥前馈神经网络可以任意逼近平方可积函数的优势,对传统事件触发控制过程进行学习,有效解决了传统方法对具体数学模型的依赖问题,且可进一步降低事件触发控制的次数。另外,相较于传统的整数阶控制器,分数阶控制器拓宽了事件触发控制算法的实际应用范围。以具体应用为背景进行仿真分析,结果表明,该算法不但能够保持传统事件触发控制算法在节省控制资源方面的优势,而且在可移植性方面优于传统的事件触发控制算法。To improve the universality of event-triggered control algorithms and reduce the dependence on specific mathematical expressions,we propose a feedforward neural network-based event-triggered control method,leading to a more generalized design of fractional-order event-triggered controllers.This method effectively utilizes the ability of feedforward neural networks to approximate square integrable functions,learning from the traditional event-triggered control processes,and effectively resolves the dependency issues on specific mathematical models in traditional methods,while further reducing the frequency of event-triggered control.In addition,compared with the traditional integer-order controller,the fractional-order controller extends the practical application of event-triggered control algorithms.Simulation analysis,conducted in the context of specific applications,demonstrates that this algorithm not only maintains the advantages of traditional event-triggered control algorithms in saving control resources but also surpasses them in terms of portability.

关 键 词：前馈神经网络 数据驱动 事件触发控制 分数阶系统 同步控制 

分 类 号：TN332[电子电信—物理电子学]