基于改进PSO-BP神经网络的网络控制系统时延预测  

Delay prediction of network control system based onimproved PSO-BP neural network

作　　者：魏天旭 赵燕成 赵景波 胡阵 WEI Tian-xu;ZHAO Yan-cheng;ZHAO Jing-bo;HU Zhen(School of Information and Control Engineering,Qingdao University of Technology,Qingdao 266520,China)

机构地区：[1]青岛理工大学信息与控制工程学院,山东青岛266520

出　　处：《陕西科技大学学报》2024年第3期158-165,173,共9页Journal of Shaanxi University of Science & Technology

基　　金：国家自然科学基金项目(51475251);山东省重点研发计划项目(2023RZA02017);山东省青岛市民生计划项目(22-3-7-xdny-18-nsh)。

摘　　要：针对网络控制系统存在的随机时延问题,本文基于BP神经网络(Back Propagation Neural Network, BPNN)建模方法,在PSO(Particle Swarm Optimization)算法的基础上引入遗传算法中交叉和变异的思想,同时对惯性权重和学习因子采用线性递减和异步时变的改进策略,提出了一种性能更优的改进PSO算法,并用该算法优化BP神经网络,构建了一种改进PSO-BP神经网络的时延预测模型;然后运用MATLAB TrueTime2.0工具箱搭建仿真平台,结合获取到的历史时延采样数据对改进PSO-BP时延预测模型和PSO-BP、BP模型进行性能对比测试.实验表明本文所提出模型的预测精度更高,误差更小,能较好的解决网络控制系统的随机时延预测问题.To address the problem of stochastic delay in network control systems,this paper introduces the idea of crossover and variation in genetic algorithm based on the PSO algorithm,and uses a linear decreasing and asynchronous time-varying improvement strategy for inertia weights and The improved PSO algorithm with better performance is proposed,and the BP neural network is optimized with this algorithm to construct an improved PSO-BP neural network delay prediction model;then the MATLAB TrueTime2.0 toolbox is used to build a simulation platform and combine the obtained historical delay sampling data to improve the PSO-BP neural network delay prediction model and the PSO-BP neural network delay prediction model.BP delay prediction model and PSO-BP,BP model for performance comparison test.The experiments show that the proposed model has higher prediction accuracy and smaller error,which can better solve the stochastic delay prediction problem of network control system.

关 键 词：网络控制系统 PSO算法 BP神经网络 网络诱导时延 时延预测 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]