中文核心期刊
CSCD来源期刊
中国科技核心期刊
RCCSE中国核心学术期刊

重庆交通大学学报(自然科学版) ›› 2020, Vol. 39 ›› Issue (09): 8-16.DOI: 10.3969/j.issn.1674-0696.2020.09.02

• 交通+大数据人工智能 • 上一篇    下一篇

基于时空特性和灰色残差的交通故障数据诊断与修复

陆百川1,2,张冬梅1,舒芹1,李玉莲1   

  1. (1. 重庆交通大学 交通运输学院,重庆 400074; 2. 重庆交通大学 重庆山地城市交通系统与安全实验室,重庆 400074)
  • 收稿日期:2019-04-17 修回日期:2019-06-19 出版日期:2020-09-18 发布日期:2020-09-22
  • 作者简介:陆百川(1961—),男,江苏南通人,教授,主要从事智能交通、交通信息工程及控制方面的研究。E-mail:656542576@qq.com 通信作者:张冬梅(1995—),女,四川乐山人,硕士研究生,主要从事交通信息工程及控制方面的研究。E-mail:838292401@qq.com
  • 基金资助:
    中国博士后科学基金面上项目(2016M592645)

Diagnosis and Repair of Traffic Fault Data Based on Spatio-Temporal Characteristics and Grey Residual

LU Baichuan1,2, ZHANG Dongmei1, SHU Qin1, LI Yulian1   

  1. (1. School of Traffic &Transportation, Chongqing Jiaotong University, Chongqing 400074, China; 2. Key Laboratory of Traffic System & Safety in Mountain Cities, Chongqing Jiaotong University, Chongqing 400074, China)
  • Received:2019-04-17 Revised:2019-06-19 Online:2020-09-18 Published:2020-09-22

摘要: 针对交通检测器获取的数据可能存在不完整或异常等问题,同时路网中交通流数据之间存在时间相关和空间相关的特征,笔者提出了基于时空特性和灰色残差的交通故障数据诊断与修复方法。在分析故障数据特点基础上考虑交通流数据存在的时间、空间特性,建立基于马氏距离(MD)的故障数据诊断模型;然后利用灰色关联模型计算出时空维度相关系数,选择强关联路段时空相关数据建立基于灰色—遗传小波神经网络(GM(1,N)-GA-WNN)的故障数据修复模型。研究结果表明:基于GM(1,N)-GA-WNN模型的故障数据诊断与修复的平均相对误差与GM-WNN模型相比降低了1.7%,与GM(1,N)模型相比降低了4.8%;考虑时空特性的故障数据修复精度优于单一时间、空间修复序列,为道路交通预测、交通诱导等服务提供可靠的数据保障。

关键词: 交通工程, 交通流, 灰色关联分析, 故障诊断, 数据修复, 时空特性

Abstract: The data acquired by traffic detectors may be incomplete or abnormal, and there were temporal and spatial correlations between traffic flow data in the road network, so the traffic fault data diagnosis and repair method based on spatio-temporal characteristics and grey residuals was proposed. Firstly, on the basis of analyzing the characteristics of fault data, the temporal and spatial characteristics of traffic flow data were considered, a fault data diagnosis model based on Mahalanobis distance (MD) was established. Secondly, the correlation coefficients of spatial-temporal dimension were calculated by using grey correlation model, and the time-spatial correlation data of strongly correlated sections were selected to establish the fault data repair model based on grey-genetic wavelet neural network (GM(1,N)-GA-WNN). The case study shows that the average relative error of fault data diagnosis and repair based on GM(1,N)-GA-WNN model is 1.7% lower than that of GM-WNN model and 4.8% lower than that of GM (1, N) model. Meanwhile, the accuracy of fault data repair considering spatio-temporal characteristics is better than that of single time and space repair sequence, which provides reliable data guarantee for road traffic prediction, traffic guidance and other services.

Key words: traffic engineering, traffic flow, grey correlation analysis, fault diagnosis, data repair, spatio-temporal characteristics

中图分类号: