交通流量VNNTF神经网络模型多步预测研究

发布时间：2018-02-25 17:02

本文关键词： 相空间重构泛函级数多步预测 VNN神经网络算法混沌　出处：《自动化学报》2014年09期 　论文类型：期刊论文

【摘要】：研究了VNNTF神经网络(Volterra neural network traffic flow model,VNNTF)交通流量混沌时间序列多步预测问题.通过分析比较交通流量混沌时间序列相空间重构的嵌入维数和Volterra离散模型之间的关系,给出了确定交通流量Volterra级数模型截断阶数和截断项数的方法,并在此基础上建立了VNNTF神经网络交通流量时间序列模型;设计了交通流量Volterra神经网络的快速学习算法;最后,利用交通流量混沌时间序列对VNNTF网络模型,Volterra预测滤波器和BP网络进行了多步预测实验,比较了多步预测结果的仿真图、绝对误差的柱状图以及归一化后的方均根;实验结果表明VNNTF神经网络的多步预测性能明显优于Volterra预测滤波器和BP神经网络.
[Abstract]:The multistep prediction of traffic flow chaotic time series based on VNNTF neural network traffic flow model is studied. The relationship between the embedded dimension of phase space reconstruction of traffic flow chaotic time series and the Volterra discrete model is analyzed and compared. The method of determining the truncation order and the number of truncation terms of Volterra series model of traffic flow is given. On the basis of this, the traffic flow time series model of VNNTF neural network is established, the fast learning algorithm of Volterra neural network of traffic flow is designed, and the fast learning algorithm of Volterra neural network of traffic flow is designed. Using chaotic time series of traffic flow, the multistep prediction experiment of VNNTF network model is carried out, and the simulation diagram of multi-step prediction results, the histogram of absolute error and the normalized square mean root are compared. The experimental results show that the multi-step prediction of VNNTF neural network is better than that of Volterra predictive filter and BP neural network.
【作者单位】：合肥工业大学电气与自动化工程学院;
【基金】：国家杰出青年科学基金(50925727) 教育部科学技术研究重大项目(313018) 安徽省高校自然科学基金重点项目(KJ2012A219) 中国博士后科学基金(2013M541823)资助~~
【分类号】：U491.1;TP183

【相似文献】