高速公路合流区上下游交通流量特性分析及预测研究

发布时间：2018-11-13 12:16

【摘要】：高速公路合流区是高速公路的“瓶颈”路段。进入合流区的车辆主要是上游主线车流量和相连入口匝道车流量，两部分车辆合流时，由于主线和匝道流量变化、速度不一致，会引起主线交通流紊乱，易导致安全问题发生，使得通行效率下降。由于合流区的流量与主线和匝道流量具有相关性，通过分析合流区上下游的流量特性，掌握合流区的流量变化规律，判断其发展的趋势，对保证车辆畅通运行，提高合流区的通行效率和交通安全具有重要的意义。为此，本文依据高速公路合流区上下游检测器获取的交通流数据，分析合流区的流量与主线和匝道流量的时空相关性，并将时空关联分析与支持向量回归机结合用于实现多断面影响下高速公路合流区流量的短时预测，研究建立了适合高速公路合流区短时流量预测的回归预测模型。主要研究内容包括： ①高速公路合流区上下游交通流量特性分析。首先，将高速公路合流区交通数据时间序列分为环比时间序列和同比时间序列；然后，引入相似性度量函数来测度时间关联性和空间关联性，同时，通过“分时段”的相似性测度来对比分析高速公路合流区交通流量变化特性；最后，利用重庆渝武高速公路G75中环快速干道互通立交合流区的实测数据进行分析与验证。 ②基于时空关联性分析与支持向量机回归结合的预测模型的建立。首先，，分析传统采用相邻前n个时间段的交通流数据作为输入建立的支持向量机回归（SVR）预测模型的不足；然后，利用合流区时空关联分析结果，改进了传统支持向量机回归模型，建立了基于时空关联性分析与支持向量机回归结合的合流区流量预测模型；最后，通过网格搜索、遗传算法和粒子群算法来获取SVR的参数，并利用实测数据来分析改进模型的预测效果。 ③合流区流量时序峰值预测的加权最小二乘支持向量机回归模型的建立。首先，针对流量时序峰值样本拟合预测误差偏大的问题，基于已有研究成果并结合加权最小二乘思想，设计了基于自信息的拟合误差加权修正系数；然后，利用设计的拟合误差加权修正系数来增大峰值样本拟合误差权重来实现对流量时序峰值拟合回归预测；最后，利用重庆市渝武高速公路的实测流量峰值数据对模型进行了分析和验证。
[Abstract]:The confluence zone of expressway is the bottleneck section of expressway. The vehicles entering the confluence zone are mainly the upstream main line traffic flow and the connected on-ramp traffic flow. When the two parts of the vehicle converge, due to the variation of the main line and ramp flow, the speed is inconsistent, which will cause the main line traffic flow disorder, which will easily lead to safety problems. Reduce the efficiency of traffic. Because the flow of the confluence zone is related to the main line and the ramp flow, by analyzing the flow characteristics of the upstream and downstream of the confluence zone, we can master the law of the flow change in the confluence zone, judge its developing trend, and ensure the smooth operation of the vehicle. It is of great significance to improve the traffic efficiency and traffic safety in the confluence area. Therefore, based on the traffic flow data obtained by the upstream and downstream detectors in the freeway confluence area, this paper analyzes the space-time correlation between the traffic flow in the confluence area and the main line and ramp flow. The combination of time and space connection analysis and support vector regression machine is used to realize the short time prediction of the flow in the confluence area of freeway under the influence of multi-section, and a regression forecasting model suitable for the prediction of short time flow in the confluence area of expressway is established. The main research contents are as follows: 1 Analysis of traffic flow characteristics of upstream and downstream in freeway confluence area. First, the time series of traffic data in freeway confluence area is divided into the time series of ring comparison and the time series of year on year. Then, the similarity measure function is introduced to measure the temporal and spatial correlation. At the same time, the characteristics of the traffic flow in the freeway confluence area are compared and analyzed by the similarity measure of "divided time". Finally, the measured data of G75 Central Expressway Interchange and Interchange area of Chongqing Yu-Wu Expressway are analyzed and verified. 2 the establishment of prediction model based on the combination of spatiotemporal correlation analysis and support vector machine regression. Firstly, the shortcomings of the traditional support vector machine (SVM) regression (SVR) prediction model based on the traffic flow data of the first n adjacent time periods are analyzed. Then, the traditional support vector machine regression model is improved, and the flow prediction model based on the combination of spatio-temporal correlation analysis and support vector machine regression is established. Finally, the parameters of SVR are obtained by grid search, genetic algorithm and particle swarm optimization algorithm, and the prediction effect of the improved model is analyzed by using the measured data. (3) the establishment of weighted least squares support vector machine regression model for predicting the peak value of flow time series in the confluence region. Firstly, aiming at the problem that the prediction error of peak sample fitting of flow time series is too large, based on the existing research results and the idea of weighted least squares, the weighted correction coefficient of fitting error based on self-information is designed. Then, the weight of the peak sample fitting error is increased by using the weighted correction coefficient of the designed fitting error to realize the prediction of the peak fitting regression of the flow time series. Finally, the model is analyzed and verified by using the measured peak flow data of Chongqing Yuwu Expressway.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491.1

【参考文献】