高速公路短时交通量预测

发布时间：2018-09-06 18:59

【摘要】：随着高速公路的不断发展与建设,由于其高效快速的通行能力已被越来越多的人作为日常出行方式的主要选择。但是,随着人们的出行需求的不断提高,已有的路网通行能力早已不能满足现有出行量,随即产生越来越多的交通拥堵状况,而其后果是交通事故的频发与环境污染的恶化。因此,对高速公路短时交通量预测理论和方法的研究是实现合理有效的交通诱导、缓解交通拥堵,减少交通事故的频发以及改善环境污染的必要前提。当今高速公路数据采集技术与设备的不断完善,使得高速公路短时交通量的预测成为可能,本文立足于高速公路短时交通量的可预测性,建立并改进了不同的预测模型,并对其预测结果进行了比较。首先,本文对短时交通量预测的研究背景、意义以及国内外研究现状进行了总结,分析了各类预测方法和模型中存在的不足,应用有效的采集手段统计了兰海高速与武罐高速两条高速公路的短时交通流量用于后续模型的实际分析中。其次,针对高速公路短时交通量时间序列的内部特性,介绍了混沌理论的一些基本概念和参数,为了更好地分析该时间序列,重构其相空间,通过C-C算法计算兰海高速和武罐高速的两组实验数据的延迟时间和嵌入维数,重构了原始时间序列的相空间,以此将其内部存在的实际规律挖掘出来,在此基础上,利用小数据量法计算两组数据的最大李雅普诺夫指数,计算结果均大于零,表明两组数据都可以利用混沌理论对其进行相应的分析与研究。然后,介绍人工神经网络的相关概念,利用小波神经网络和RBF神经网络预测高速公路短时交通量,在此之前,分别利用重构的两组数据的延迟时间和嵌入维数合理设计网络的输入层与输出层神经元的数量,建立良好的网络拓扑结构,对兰海高速和武罐高速采集的数据在建立的网络中进行预测实验,通过分析计算实验结果可以得出,RBF神经网络的预测效果比小波神经网络更好。最后,针对两种神经网络中存在的不足之处,使用遗传算法对两种网络的初始参数进行最优选择,以保证网络的输出结果更加良好,在对兰海高速和武罐高速采集的两组数据进行预测实验后,可以得出改进后的两种神经网络预测误差均得到了改善,同时,改进后的RBF神经网络预测模型也优于改进后的小波神经网络预测模型,可以更好地实现对高速公路短时交通量的预测。
[Abstract]:With the continuous development and construction of freeway, more and more people have taken it as the main choice of daily travel mode because of its efficient and rapid traffic capacity. However, with the continuous improvement of people's travel demand, the existing road network capacity has been unable to meet the existing travel volume, resulting in more and more traffic congestion, and the consequences are the frequent occurrence of traffic accidents and the deterioration of environmental pollution. Therefore, the research on the theory and method of short-term traffic volume prediction is a necessary prerequisite for realizing reasonable and effective traffic guidance, alleviating traffic congestion, reducing frequent traffic accidents and improving environmental pollution. Nowadays, with the continuous improvement of data acquisition technology and equipment of expressway, it is possible to predict the short-term traffic volume of expressway. Based on the predictability of short-term traffic volume of expressway, different prediction models are established and improved. The predicted results are compared. First of all, this paper summarizes the research background, significance and current research situation of short-term traffic volume prediction, and analyzes the shortcomings of various forecasting methods and models. In this paper, the short time traffic flow of Lanhai high speed highway and Wucan high speed highway is analyzed by means of effective collection method. Secondly, some basic concepts and parameters of chaos theory are introduced in order to better analyze the time series and reconstruct its phase space. The delay time and embedding dimension of two groups of experimental data of Lanhai high-speed and Wu-can high-speed are calculated by C-C algorithm, and the phase space of the original time series is reconstructed. The maximum Lyapunov exponent of the two groups of data is calculated by the method of small amount of data, and the results are all greater than zero, which indicates that both groups of data can be analyzed and studied by chaos theory. Then, the related concepts of artificial neural network are introduced. Wavelet neural network and RBF neural network are used to predict the short-time traffic volume of freeway. By using the delay time and embedding dimension of the reconstructed two groups of data, the number of neurons in the input and output layers of the network is reasonably designed, and a good network topology is established. The prediction experiments of the data collected at the high speed of Lanhai and Wu-can are carried out in the established network. By analyzing and calculating the experimental results, it can be concluded that the prediction effect of RBF neural network is better than that of wavelet neural network. Finally, aiming at the shortcomings of the two neural networks, genetic algorithm is used to optimize the initial parameters of the two networks to ensure that the output of the network is better. After the two groups of data collected at the high speed of Lanhai and Wu-can are forecasted, it can be concluded that the two improved neural networks have improved the prediction error, and at the same time, The improved RBF neural network prediction model is also better than the improved wavelet neural network prediction model, which can better realize the short-term traffic volume prediction of expressway.
【学位授予单位】：兰州交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491.14

【参考文献】