【作者】 张杰；

【导师】 赖元文；

【作者基本信息】 福州大学 ， 交通运输工程（专业学位）， 2020， 硕士

【摘要】 公交调度作为公交运营管理的重要手段,公交调度的合理性对公交资源利用及乘客满意程度有着较大影响,公交客流特征可为公交调度提供信息支撑。随着智能公交技术的发展,公交刷卡数据及公交GPS数据等公交多源数据的获取变得更加便捷、准确。以往的公交调度方法在公交多源数据的获取与挖掘上存在一定的局限。因此,有必要对公交多源数据进行深入挖掘,识别潜在的线路客流特征,结合客流特征对公交调度优化方法进行研究,以优化公交运营资源、提高乘客满意程度,从而增强公交出行的吸引力。本文以常规公交为研究对象。首先,根据公交多源数据的特点,分析数据预处理技术。在对原始数据进行数据清洗的基础上,针对公交刷卡系统及公交GPS系统二者数据上传可能存在的时间误差,运用系统误差修复方法对二者误差进行校正;针对公交GPS数据存在到离站间数据缺失的问题,运用最大概率模型对缺失站间数据进行修复,以实际公交运营数据为例验证了模型的可行性。其次,提出公交线路客流特征识别方法。一是考虑刷卡数据具有严格时间排序性,提出Fisher最优分割法将具有相似性的刷卡数据聚类,划分客流区间。二是针对刷卡数据缺陷,提出了乘客刷卡时间与公交到离站时间相匹配的上车站点识别方法。三是将乘客按照是否具有出行规律分为两个部分。一部分是在工作日出行具有出行规律的乘客,提出了基于乘客出行规律的乘客下车站点识别方法;另一部分是单程、换乘出行或偶然出行的不具有出行规律的乘客,为提高这部分乘客的识别精度,提出包含出行距离、换乘能力、出行吸引强度、周围土地利用性质等因素影响下的乘客站点下车概率识别方法。再次,对考虑客流特征的公交调度模型及求解算法进行优化设计。一是结合客流特征,建立了综合公交公司和乘客双方利益的公交调度优化模型。二是以新兴仿生群算法布谷鸟算法为基础,改进其缺陷,并通过测试函数将改进算法与传统布谷鸟算法、粒子群算法进行寻优对比,结果表明,改进算法的寻优性能最佳。最后,以福州125路公交为例,验证所提的公交线路客流特征识别方法及公交调度优化方法的有效性。对该线路公交数据进行挖掘,识别客流特征,通过与实际调查结果对比与误差分析,验证客流特征识别方法的准确性。将该客流特征分析结果应用于模型和求解算法中,结果表明,基于不同利益方权重下通过模型算法计算出的结果均优于现有调度方案,验证了模型及算法的有效性及实用性。更多还原

【Abstract】 As an important means of bus operation management,the rationality of bus scheduling has a great influence on the utilization of bus resources and the satisfaction of passengers.The characteristics of bus passenger flow can provide information support for bus dispatching.With the development of intelligent bus technology,the acquisition of bus multi-source data such as bus payment data and bus GPS data becomes more convenient and accurate.The previous bus scheduling methods have some limitations in the acquisition and mining of bus multi-source data.Therefore,it is necessary to conduct in-depth mining of bus multi-source data,identify the potential characteristics of passenger flow,and study the optimization method of bus scheduling based on the characteristics of passenger flow,so as to optimize the operating resources of bus and improve the satisfaction of passengers,thus enhance the attractiveness of bus travel.This paper takes the conventional public transit as the research object.Firstly,according to the characteristics of bus multi-source data,the data preprocessing technology is analyzed.Based on the data cleaning of the original data,the paper corrects the possible time errors of the bus payment system and the bus GPS system with the method of system error repair.Aiming at the problem of missing data between bus arrival time data and leave time data,the maximum probability model was used to repair the missing data between stations,and the feasibility of the model was verified by taking the actual bus operation data as an example.Secondly,the paper proposes a method to identify the passenger flow characteristics of bus lines.First,considering the strict time ordering of swipe data,the Fisher optimal segmentation method is proposed to cluster similar payment data and divide the passenger flow interval.Second,aiming at the defects of bus payment data,the paper proposes a method to identify the boarding station which matches the time of passengers’ card payment with the time of buses’ arrival and leave.Third,passengers are divided into two parts according to whether they have travel rules.The first part is about the passengers who travel regularly on weekdays,and the identification method of passenger alighting station based on the rule of passenger travel is proposed.In order to improve the identification accuracy of these passengers,a probabilistic identification method for passenger station alighting under the influence of travel distance,transfer ability,travel attraction intensity and surrounding land use property is proposed.Thirdly,the optimal design of the bus scheduling model and algorithm considering the characteristics of passenger flow is carried out.First,based on the characteristics of passenger flow,a bus scheduling optimization model is established to integrate the interests of both bus companies and passengers.Second,based on the Cuckoo Search algorithm of the new bionic swarm algorithm,the defects of the improved algorithm are improved,and the improved algorithm is compared with the traditional Cuckoo Search algorithm and Particle Swarm Optimization algorithm through the test function.Finally,taking Fuzhou 125 bus line as an example,the effectiveness of the proposed method of passenger flow feature identification and bus scheduling optimization is verified.By mining the bus data of the line and identifying the characteristics of passenger flow,the accuracy of the identification method of passenger flow characteristics is verified by comparing with the actual survey results and error analysis.The analysis results of the passenger flow characteristics are applied to the model and algorithm,and the results show that the results calculated by the model algorithm based on the weights of different stakeholders are better than the existing scheduling scheme,which verifies the effectiveness and practicability of the model and algorithm.更多还原