基于广义出行成本的交通方式划分与随机分配组合模型研究

发布时间：2018-05-12 10:10

本文选题：交通需求预测 + 方式划分　；参考：《吉林大学》2015年硕士论文

【摘要】：随着我国城镇化和机动化出行水平的提高，城市交通拥堵状况不断加剧。交通拥堵以及由此引发的交通事故和环境污染问题已经影响到城市生活的效率和质量。在拥堵治理策略上，多数国家已转向以管理交通需求为主，增加交通供给为辅。但无论何种策略，在实施前，模拟和预测策略实施效果、评价策略可行性都是十分必要的。其中，方案效果评价的两个重要指标是预测实施后的各交通方式分担率和交通网络分配流量。传统的“四阶段”法将交通方式划分和交通分配分为两个独立的阶段分别进行预测，，弱化了两者之间的密切联系和相互依赖关系，存在一些明显不合理的地方，降低了预测的准确性。因此，十分有必要研究交通方式划分与分配组合模型，以提高预测精度，为交通行为分析与研究提供理论支持，为交通规划方案、管理政策的评估提供依据。为实现这一目的，首先采用实际调查数据对影响居民出行选择的三方面因素（出行主体特性、出行特性和交通方式特性）的影响情况逐一进行了分析，并根据影响因素作用机理的不同进行分类，筛选出主要影响因素；其次，提出了城市交通超级网络的构造方法，描述了不同子网络层应具有的属性值，以此为基础，将各交通方式出行成本分解为出行时间、费用、舒适度三个方面，逐一讨论了各方式广义出行成本的表征；此后，给出了组合模型的基本假设、自变量、因变量和约束条件，引用Logit随机平衡理论给出了方式划分与分配组合模型的平衡条件，构造了变分不等式模型，并证明了模型的等价性及解的存在性，给出了具体求解步骤；最后，以山东省中北部G县为例进行了实例分析，阐述了模型的具体应用方法，并将组合模型预测结果、传统分阶段模型预测结果及实际调查数据进行比较，分析了模型的优劣。结果表明，在交通方式划分方面，组合模型累计预测误差比Logit模型低13.2%；在道路断面车流量预测方面，组合模型的预测误差为15.32%，比随机用户平衡模型降低了6.35%；在公交站间断面客流量预测方面，组合模型的预测误差为18.05%，比随机用户平衡模型降低了4.36%。组合模型在三个方面的预测精度都有一定程度的提高，具有良好的应用价值。
[Abstract]:With the improvement of urbanization and motorized travel level in China, traffic congestion in cities is increasing. Traffic congestion and the resulting traffic accidents and environmental pollution have affected the efficiency and quality of urban life. In the strategy of congestion control, most countries have turned to the management of traffic demand and increase the supply of traffic. But no matter what kind of strategy, it is necessary to simulate and predict the implementation effect of the strategy and evaluate the feasibility of the strategy before it is implemented. Among them, the two important indexes of the project effect evaluation are the traffic mode sharing rate and the traffic network distribution flow after the implementation. The traditional "four stage" method divides the traffic mode and the traffic distribution. It is divided into two independent stages, which weaken the close relations and interdependence between the two. There are some obvious and unreasonable places, which reduce the accuracy of the prediction. Therefore, it is necessary to study the model of traffic mode division and distribution combination to raise the accuracy of prediction and provide a rational analysis and research for traffic behavior. Support is the basis for traffic planning and management policy evaluation.
In order to achieve this goal, first of all, the actual survey data are used to analyze the influence of three factors (travel characteristics, travel characteristics and traffic characteristics) on the travel choice of residents, and the main influencing factors are selected according to the different mechanism of influence factors. Secondly, the urban intersection is put forward. According to the construction method of super network, the attribute values of different sub network layers should be described. Based on this, the travel cost is decomposed into three aspects of travel time, cost and comfort, and the characterization of the generalized travel cost is discussed one by one. After that, the basic assumptions, variables, variables, and variables of the combined model are given. Constraint conditions, the equilibrium condition of mode division and distribution combination model is given by using Logit stochastic equilibrium theory. A variational inequality model is constructed. The equivalence of the model and the existence of the solution are proved. The concrete solution is given in the case of G County in the middle north part of Shandong Province, and the concrete needs of the model are expounded. By using the method, the prediction results of the combined model, the traditional model and the actual survey data are compared, and the advantages and disadvantages of the model are analyzed.
The results show that the cumulative prediction error of the combined model is 13.2% lower than that of the Logit model, and the prediction error of the combined model is 15.32%, which is 6.35% lower than that of the random user equilibrium model. The prediction error of the combined model is 18.05%, and the prediction error of the combined model is 18.05%, and the ratio of the combined model is 18.05%. The stochastic user equilibrium model reduces the prediction accuracy of the 4.36%. combination model in three aspects to a certain extent, and has good application value.

【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491.1

【参考文献】