杨泽林
入学时间：2017级
答辩时间：2020年
论文题目：考虑韧性指标的公共汽车线路运行可靠度评价方法研究
中文摘要
摘要
随着我国城镇化进程的加快，越来越多的城市面临严重的交通拥堵问题，通过优先发展公共交通来满足城市居民日益增长的机动化出行需求已经成为政府部门和社会各界的共识。提高公交服务水平的基础是要提高公交线路的运行可靠度，亦即公交线路在一定时间内、在一定条件下能够以规定的标准提供运输服务的能力或可能性。传统公交线路的可靠度评价方法存在指标体系不全、定量分析不足、方法通用性差等问题，尤其是忽略了线路从干扰中自行恢复的能力即线路韧性的影响。因此本文旨在优化已有评价方法的同时考虑韧性指标对可靠度的影响。鉴于韧性和可靠度是揭示公交线路属性的两种维度，文章分别建立模型对二者进行评价，进一步结合数据分析韧性和可靠度的关系。综上，考虑韧性指标的公共汽车线路运行可靠度研究具有重要的理论价值与现实意义。
首先，本研究对运行可靠度的影响因素进行分析，建立由运行速度可靠度、发车间隔可靠度和准点可靠度等指标组成的评价体系，进一步对上海市10条线路的自动车辆位（AVL）数据、中间站时刻表等相关数据进行融合挖掘，在此基础上建立了适用于公交线路运行可靠度的效能评价（ADC）模型。其次，从韧性的基本定义出发，结合韧性三角形的确定方法，建立公交线路韧性量化模型，选择最大恢复韧性（RRIR）、均匀韧性（MRIR）、区间加权韧性（WRIR）三个指标作为韧性研究的切入点，给出每种指标基于AVL数据的计算方法。最后，选择典型相关分析法进一步探究韧性和可靠度的关系，计算得出典型相关系数CR1为0.739即证明韧性与可靠度存在较为明显的相关关系，且韧性越高的线路可靠度越高。典型相关系数结构图证明线路最大恢复韧性对可靠度的影响十分显著，提高线路最大恢复韧性可使运行速度可靠度和准点可靠度两个指标得到明显地改善，进而显著提高线路的运行可靠度。
本文在总结国内外公交线路可靠度评价相关研究的基础上，考虑公交线路韧性指标，并着重建模探究了韧性和可靠度的关系，较为全面地解析公交运行可靠度的内涵，在完善已有评价方法缺点的基础上回归现实世界，分析了如线路长度、公交专用道里程等因素对可靠度和韧性的影响，研究结论有助于公交企业合理优化线路，提升公交服务品质，同时可以为政府补贴公交企业提供依据。
关键词：公交线路；可靠度；韧性；AVL数据
英文摘要
ABSTRACT
With the process of urbanization, more and more cities are facing serious traffic congestion problems,which must be solved by developing public transportation. It is very important to improve the operational reliability of bus lines. That is, the ability or possibility that the bus line can provide transportation services within a certain period of time under the certain conditions.The reliability evaluation methods of bus lines have problems such as incomplete index system, insufficient quantitative analysis, and poor method versatility, especially ignoring the influence of the line's ability to recover from interference, which is toughness.Therefore, this study aims to optimize existing evaluation methods while considering the impact of toughness indicators on reliability.In summary, the study of the reliability of bus line operation considering the toughness index has important theoretical and practical significance.
First and foremost, this study analyzes the influencing factors of operating reliability, and establishes an evaluation system consisting of operating speed reliability, departure interval reliability, and on-time reliability.Immediately afterwards, the relevant data such as the automatic vehicle position (AVL) data of the 10 lines in Shanghai and the timetable of the intermediate station were merged and mined. On this basis, an effectiveness evaluation (ADC) model suitable for the operation reliability of bus lines is established.Secondly, starting from the basic definition of toughness, combined with the determination method of the toughness triangle, a quantitative model of bus line toughness was established, and three indexes recover resilience index based on route (RRIR), mean resilience index based on route (MRIR), Weighted average resilience index based on route (WRIR) were selected as the toughness research The entry point for each indicator is given based on the calculation method of AVL data.Finally, in order to further explore the relationship between reliability and toughness, select the canonical correlation analysis method to discuss the relationship between toughness and reliability. The calculated typical correlation coefficient is 0.739, which proves that there is a relatively obvious correlation between toughness and reliability.The typical correlation coefficient structure diagram proves that the maximum recovery toughness of the line has a significant effect on reliability.
Based on the summary of relevant research on the reliability evaluation of bus lines at home and abroad, this paper considers the toughness indexes of bus lines. And focused on modeling to explore the relationship between toughness and reliability. The conclusion of this study is helpful for the bus to rationally optimize the bus route and improve the quality of bus service.
Key words:Bus routes; Reliability; Resilience; AVL data
康振
入学时间：2017级
答辩时间：2020年
论文题目：基于数据的城市道路间断流流量预测方法研究
中文摘要
摘要
交通流量描述了通过特定断面的车辆数，是反映交通状况的重要信息。对交通流量的预测，一般分为长时、中时、短时交通流预测等三种。长时预测的周期一般在几天乃至数月之间，中时预测一般对周期数个小时之间的交通流状况进行预测，短时预测一般对1分钟到15分钟之间的交通流状况进行预测。长时、中时交通流量预测服务于道路规划建设与交通管理，已有较成熟的研究成果与应用模式。短时交通流量预测是实现智能交通控制和诱导的关键环节，也是提高现有交通路网的通行效率的有效手段，还是对数据进行深度挖掘、完善交通理论的重要方向。近年来，在ITS的影响下，短时流量预测的应用越来越引起管理者与研究者的重视。
大量流量预测研究基于交通流的时间序列，提出了多种模型进行短时交通流量预测，并在高速公路、快速路等连续流道路中获得了较好的应用。但是，当模型迁移至城市道路中时，由于城市道路交通流受到交叉口信号控制、交通需求的动态性和不确定性及大量交通事件的影响而呈现出剧烈波动性和强随机性特点，上述预测模型不能够很好的契合交通流量的变化特性，也不能够较好的规避随机因素对流量的预测产生的影响，存在模型的鲁棒性不好、预测精度不够高、应用效果差等问题。
为了将短时流量预测更好的应用在城市道路中，部分研究者通过训练复杂模型对城市道路的随机性、波动性进行拟合，从而得到较高的预测精度，但存在模型复杂度过高，训练时间长而落地应用难等缺陷，不符合工程应用需求。同时，城市道路中广泛存在有路段内出入口流入流出车辆，现有的研究无法解释该类车辆的出行情况，导致预测结果存在固定误差，且该类固定误差在高峰时刻尤为突出，造成很大的精度影响。因此，城市道路流量预测亟需寻找新的突破口。
随着ITS的逐步实施，各类交通检测设备提供了不同精度、广度和深度的大量交通数据。传统预测方法往往具有特定的模型结构和较强的假设条件，只需对数据进行集计处理后就进入模型训练，在此过程中，忽略了高质量高精度数据的非集计特征，而该类非集计特征能够有效帮助车辆进行溯源预测，对城市道路间断流流量预测的精度有重要影响。因此，有必要研究和探索用于城市道路间断流交通流量预测的新方法，以充分挖掘交通数据所蕴含的丰富交通信息，进一步提升预测的准确性、可靠性与落地应用价值。
基于上述考虑，本文旨在建立适用于城市道路间断流场景下的流量预测方法，主要研究内容如下：
1)从短时交通流预测方法对国内外研究现状进行了总结，提出在城市道路流量预测中应考虑上下游路段、考虑数据的非集计特征、考虑路段出入口流入流出流量、考虑交通流参数之间的内在关联等因素的影响。
2)对比分析现有交通采集技术的发展现状与数据质量，并从交通数据的集计特征与非集计特征进行数据比较研究。从对比结果来看，城市道路卡口数据是适用于城市道路流量预测的主要数据源，其集计特征与非集计特征能够有效结合进行车辆溯源，进而对道路流量进行预测。
3)提出基于卡口数据的城市道路间断流流量预测场景与方法。考虑卡口检测设备的分布情况，场景设计分为上游交叉口有卡口监控与上游交叉口无卡口监控的两种场景。通过建立了基础研究场景，提出基于上游流量计算的流量预测方法。
4)针对本文的流量计算方法，解决了三个关键问题：上游至下游的行程时间计算问题、上游交叉口流入比例预测问题、路段内流出流量问题。行程时间计算考虑到间断流对车辆造成的延误问题。上游交叉口流入比例预测问题与路段流出流量通过定义“流入—流出流量”的概念，简化了卡口数据的处理与计算。通过季节性ARIMA模型进行计算上游交叉口的流入比例，通过贝叶斯模型估计路段流出流量。
5)以卡口数据作为数据源进行实验，并与现有的常用方法（ARIMA模型、BP神经网络）进行对比实验，验证了本方法的预测效果与可行性。
关键词：流量预测，城市道路，间断流，数据，非集计特征，流入—流出流量，路段流出流量
英文摘要
ABSTRACT
Traffic flow describes the number of vehicles passing through a particular section and is an important information reflecting the traffic situation. According to the length of time, traffic flow prediction is classified generally divided into long-time, medium-time and short-time. Long-time traffic flow prediction is generally days or even months. Medium-time traffic flow prediction generally predicts traffic flow conditions between several hours. Short-time generally predict traffic flow between 1 minute and 15 minutes.Long-time and medium-time traffic flow prediction is used in road planning and traffic management.As for short-time traffic flow prediction, it is a key link to realize intelligent traffic control and guidance, an effective way to improve the traffic efficiency, and an important direction for data mining and traffic theory improvement. In recent years, under the influence of ITS, the application of short-time traffic prediction has attracted more and more attention from managers and researchers.
Based on the time series data of traffic flow, a variety of models are put forward for short-time traffic flow prediction, which has been well applied in continuous flow roads such as highways and expressways.However, when the model migration to urban road, because of the urban road traffic flow is intersection signal control, the dynamics and uncertainty of urban road traffic demand, and the influence of a large number of traffic incidents which present a severe volatility and randomness, the above prediction models cannot well fit the changing characteristics of traffic flow, nor can it better avoid the influence of random factors on the prediction of traffic flow. There are problems such as poor robustness, low prediction accuracy and poor application effect of these models.
In order to better apply short-time traffic prediction in urban roads, some researchers have fitted the randomness and volatility of urban roads’ traffic flow through training complex models, so as to obtain a higher prediction accuracy. However, such defects as high model complexity, long training time and difficulty in practical application do not meet engineering application requirements. At the same time, there are a wide range of vehicles flowing in and out of the road sections in urban roads, and existing studies cannot explain the travel conditions of such vehicles, resulting in a fixed error in the prediction results, which is particularly prominent in the peak hours, causing a great impact on the accuracy. Therefore, it is urgent to find a new breakthrough in urban road flow prediction.
With the gradual implementation of ITS, various kinds of traffic detection equipment provide a large number of traffic data with higher accuracy, breadth and depth. Traditional prediction methods usually have specific model structure and strong assumptions. They only need to process the data by lumpmeter and then enter into model training. In this process, the non-lumpmeter characteristics of high-quality and high-precision data are ignored. However, this kind of off-set features can effectively help vehicle traceability prediction and have an important impact on the accuracy of traffic flow prediction between urban roads. Therefore, it is necessary to study and explore a new method for the prediction of traffic flow between urban roads, so as to fully mine the rich traffic information contained in traffic data and further improve the accuracy, reliability and practical application value of the prediction.
Traditional prediction methods tend to have a specific model structure and strong assumptions, simply to disaggregate the data into the model after the training, in the process, ignore the disaggregate characteristics are of high quality and high precision data, and disaggregate characteristics are the class can effectively help the vehicle back projections, blocks the flow between the urban road traffic prediction accuracy has important influence. Therefore, it is necessary to study and explore a new method for the prediction of traffic flow between urban roads, so as to fully mine the rich traffic information contained in traffic data and further improve the accuracy, reliability and practical application value of the prediction.
Based on the above considerations, this paper aims to establish a interrupted flow prediction method suitable for the urban roads. The main research contents are as follows:
1) Based on the short-term traffic flow prediction method, this paper summarizes the current research situation at home and abroad, and proposes that the upstream and downstream sections, the non-aggregate characteristics of data, the inflow and outflow flow at the entrance and exit of sections, the internal correlation between traffic flow parameters and other factors should be considered in the urban road traffic flow prediction.
2) Compare and analyze the development status and data quality of existing traffic acquisition technologies, and conduct a data comparison study from the aggregate and non-aggregate features of traffic data. According to the comparison results, urban road bayonet data is the main data source suitable for urban road flow prediction, and its aggregate characteristics and non-aggregate characteristics can be effectively combined to carry out vehicle traceability, so as to predict the road flow.
3) The scenario and method of interrupted traffic flow prediction between urban roads based on bayonet data are proposed. Considering the distribution of bayonet detection equipment, the scene design is divided into two scenarios: upstream intersection with bayonet monitoring and upstream intersection without bayonet monitoring. The flow prediction method based on upstream flow calculation is proposed by establishing the basic research scenario.
4) Aiming at the flow calculation method in this paper, three key problems are solved: the calculation of the travel time from upstream to downstream, the prediction of the inflow proportion at the upstream intersection, and the outflow volume in the section. The calculation of travel time takes into account the problem of vehicle delay caused by intermittent flow. By defining the concept of "inflow – outflow flow", the processing and calculation of bayonet data are simplified. The inflow proportion of upstream intersections is calculated by seasonal ARIMA model, and the section outflow is estimated by bayesian model.
5) The bayonet data was used as the data source for experiments, and compared with the existing common methods (ARIMA model and BP neural network), to verify the prediction effect and feasibility of this method.
Keywords: Traffic flow prediction, Urban road, Nonaggregate feature, Inflow - outflow flow, Section outflow
