【作者】 董春娇；

【导师】 邵春福；

【作者基本信息】 北京交通大学 ， 交通运输规划与管理， 2011， 博士

【摘要】 随着交通流数据采集设备的不断完善,使得交通流状态分析、预测以及智能化交通组织成为可能,然而现有研究大多是针对单一状态下单断面的交通流短时预测,难以满足微观交通控制及网络条件下宏观交通实时诱导的需求。因此,深入研究多状态下交通流短时预测理论与方法,是提供交通信息服务,缓解道路交通拥堵,减少污染、节约能源,实现智能交通控制和实时诱导的必要理论和技术前提。本论文立足于城市快速路交通流运行的时间和空间特性,研究多状态下城市快速路网交通流短时预测模型与方法。首先,从研究思路、研究方法等方面,分类分析并综合评述国内外现有相关研究,阐述本论文的研究背景条件；其次,结合城市快速路交通流运行特点,应用线性回归分析建立了城市快速路速度——占有率模型、交通流率——占有率模型和速度——交通流率模型,明确交通流基本参数之间的关系及交通流状态划分方法,分析多状态下交通流参数时间和空间特性；再次,着眼于自由流状态、阻塞流状态和拥挤流状态下交通流参数的时间和空间特性,从交通流守恒方程和速度动态模型入手,借鉴偏微分方程组求解时空离散的思想,建立多状态下城市快速路网交通流短时预测模型,同时考虑进出口匝道、车道数变更及道路坡度等因素的影响,将交通流短时预测模型转化为交通流短时预测状态空间模型,采用定量分析与定性评价相结合的方法标定模型参数,并设计基于卡尔曼滤波方法的状态空间模型求解算法；最后分别基于交通流状态判别和交通流状态变化趋势判别的思想,结合建立的三种状态下交通流短时预测状态空间模型,构建两种混合状态下交通流短时预测方法,实现混合状态下城市快速路网交通流短时预测。本论文主要创新性研究成果如下：1、基于交通流守恒方程,构建了面向城市快速路网的交通流守恒方程组,结合自由流状态下交通流时间和空间特性,采用偏微分方程迎风格式求解的思想对其进行离散,推导出自由流状态下交通流短时预测模型,进而考虑进出口匝道、车道数变更及道路坡度等影响因素,将交通流短时预测模型转化为交通流短时预测状态空间模型,采用定量分析与定性评价相结合的方法标定了模型参数,并设计了基于卡尔曼滤波方法的模型求解算法。2、结合阻塞流状态下交通流时间和空间特性,采用Lax-Wendroff格式离散交通流守恒方程组,推导出阻塞流状态下交通流短时预测模型,同时考虑进出口匝道、车道数变更及道路坡度等影响因素,将其转化为阻塞流状态下交通流短时预测状态空间模型,标定了模型参数,并设计了基于卡尔曼滤波方法的模型求解算法。3、从速度动态模型入手,考虑拥挤流状态下交通流时间和空间特性,采用改进的Lax-Wendroff格式离散速度动态方程组,推导出拥挤流状态下交通流短时预测模型及状态空间模型,标定了模型参数,并设计了基于卡尔曼滤波方法的模型求解算法。4、构建了基于交通流状态判别和交通流状态变化趋势判别的混合状态下城市快速路网交通流短时预测方法。建立了基于SVM的道路网交通流状态判别模型和基于Elman网络的道路网交通流状态变化趋势判别模型,设计并实施了道路网交通流状态预判,采用多状态下交通流短时预测状态空间模型,实现混合状态下交通流短时预测。最后进行了实证性检验及预测结果分析。更多还原

【Abstract】 Traffic states analysis, prediction and intelligent traffic operation can be realized as the development of traffic data collection technology on the road. However, most existing researches focus on the short-term traffic flow prediction in singleness traffic state on the singleness section, which can hardly meet the needs of traffic guidance system. Therefore, it is imperative to further investigate into the methodology for short-term traffic flow prediction in multi traffic states, which is not only a core element of intelligent transportation systems but also an important base of traffic information service, traffic control and guidance which can provide travelers with efficient information and help them to choose an optimal path so as to perform path guidance, to save travel time of travelers, to relieve traffic congestion, to reduce air pollution and to save energy.This dissertation centers on the methodology for short-term traffic flow prediction in multi traffic states, taking into account the spatial-temporal characteristics of the traffic flow on urban expressway network. Firstly, the existing relevant researches were reviewed, categorized, and analyzed in perspectives of methodologies and model developments, which lays groundwork for further research in this dissertation. Secondly, established traffic flow models based on linear regression analysis with characteristics of traffic flow on urban expressway, split traffic states. An analytical model for the short-term traffic flow prediction influenced by spatial-temporal characteristics was proposed base on multi traffic states. Thirdly, accounting for the spatial-temporal characteristics in multi traffic states, such as free traffic, congested traffic and jam traffic, short-term traffic flow prediction modeles were proposed based on traffic flow conservation equation which was discrete by adopting spatial-temporal discretization idea of partial differential equation. And then, consideration the influence factors of on and off ramp, alteration of lanes number and road grade, the model of short-term traffic flow prediction was converted into the state-space model of short-term traffic flow prediction. Fourthly, the parameters of the proposed state-space model were estimated by qualitative and quantitative analysis methods and the estimation algorithm of the proposed state-space model was designed based on Kalman method. Finally, in view of identification of traffic states and identification of traffic states variation trend, two methodology of short-term traffic flow prediction were proposed based on the state-space model of short-term traffic flow prediction in multi traffic states. The short-term traffic flow prediction in multi traffic states on urban expressway network was realized by detection data.What it follows contains the main conclusions of this dissertation:1. A short-term traffic flow prediction model to consider the spatial-temporal characteristics in free traffic on urban expressway was developed based on traffic flow conservation equation. Short-term traffic flow prediction model was dispersed based on upwind difference scheme of partial differential equation. And then, consideration the influence factors of on and off ramp, alteration of lanes number and road grade, the model of short-term traffic flow prediction was converted into the state-space model of short-term traffic flow prediction in free traffic. At last, the estimation algorithm of the proposed state-space model was designed based on Kalman method.2. The model of short-term traffic flow prediction in jam traffic was proposed based on conservation equation, which was dispersed by adopting Lax-Wendroff scheme. Then, consideration the factors of on and off ramp, alteration of lanes number and road grade, the model of short-term traffic flow prediction was converted into the state-space model of short-term traffic flow prediction in jam traffic. The estimation algorithm of the proposed state-space model was designed based on Kalman method.3. Considering the spatial-temporal characteristics of traffic flow, speed model was dispersed by adopting the improved Lax-Wendroff scheme. Then, the state-space model of short-term traffic flow prediction was proposed in congested traffic. Estimation algorithm of proposed state-space model was designed based on Kalman method.4. Two models were proposed based on SVM model and Elman neural network method to realize identification of traffic states and traffic states variation trend. Short-term traffic flow prediction in multi states was realized by adopting state-space model of short-term traffic flow prediction in free traffic, congested traffic and jam traffic based on the traffic states identification. Empirical study and prediction result were analysis in the end.更多还原