【作者】 万敏；

【导师】 盛昭瀚；

【作者基本信息】 南京大学 ， 物流工程（专业学位）， 2020， 硕士

【摘要】 自2016年共享单车出现以来,其为人们提供了一种新型的绿色交通出行方式的。由于人们生活水平逐渐提高、环境保护意识逐渐增强,在面对经济飞速增长带来的一系列社会负面影响,如负荷过重的机动车使用量造成交通拥堵、环境恶化、噪声污染等,无疑使得人们对绿色、低碳出行的需求日益增加。共享单车顺势而为,不仅为低碳环保做出了一定的贡献,而且在一定程度上缓解了“最后一公里”公共交通“人的运输”的问题。但同时,由于人们在使用单车过程中产生空间转移,造成了单车无法均衡分布在各时段各区域中,从而导致利用共享单车出行会产生一系列交通问题,如人们在使用过程中无法及时找到出行单车或者乱停乱放阻碍交通等。因此,对共享单车进行系统性得合理调度是解决单车时空分布不均衡问题的重要方法。本文首先对共享单车系统进行阐述,并界定其功能需求及研究范围,基于摩拜单车微信小程序中南京市2018年10月-11月期间的出行数据,对人们使用共享单车出行的时空规律进行研究,并利用数据挖掘和可视化等相关技术,分析得出共享单车需求量在日常使用中具有高峰的特征,且高峰期持续时长大约为1小时。此外在空间位置分布上,单车通常集中于地铁、公交站点等公共场所附近。根据共享单车时空分布特征,利用层次聚类和k-means聚类对共享单车的调度区域进行划分;再分别利用BP神经网络和RBF神经网络预测模型对共享单车调度区域需求量进行预测。两种神经网络训练得出的结果都显示,预测值与实际值的接近程度较高,且通过对比,RBF模型在此类问题中的预测准确率更高,80%以上调度区域的准确率大于90%,且最高准确率达97%,模型拟合较好;接着根据共享单车调度区域的预测数据,提出合理假设,建立以成本最小化为目标、带时间窗的区域共享单车调度模型,并依据客观现实约束,提出模型的约束条件;最后通过设计遗传算法对调度模型求解,并以南京市摩拜单车真实数据为算例,求解出较合理的调度路线方案。更多还原

【Abstract】 Since the emergence of shared bicycles in 2016,it has provided people with a new way of green transportation.Due to the gradual improvement of people’s living standards and the increasing awareness of environmental protection,in the face of a series of negative social impacts brought about by rapid economic growth,such as heavy traffic caused by heavy vehicle usage,traffic congestion,environmental degradation,and noise pollution,no doubt make There is an increasing demand for green and low-carbon travel.Bicycle sharing follows the trend,not only making a certain contribution to low-carbon environmental protection,but also alleviating the "last mile" public transportation "human transportation" problem to a certain extent.But at the same time,due to the spatial transfer of people in the process of using bicycles,bicycles cannot be evenly distributed in various regions and regions,which brings a series of transportation problems to the popularity of shared bicycles.Find the bicycles in a timely manner or stop and park in disorder to obstruct the traffic,etc.Therefore,systematic and reasonable scheduling of shared bicycles is an important method to solve the problem of uneven distribution of bicycles in time and space.This article first explains the shared bicycle system,and defines its functional requirements and research scope.Based on the travel data of Mobike’s We Chat mini program in Nanjing during October-November 2018,the time and space laws for people using shared bicycle travel Carrying out research and using related technologies such as data mining and visualization,the analysis shows that the demand for shared bicycles has the characteristics of morning and evening peaks in daily use,and the duration of the peak period is about 1 hour.In addition,in terms of spatial location,bicycles are usually concentrated near public places such as subways and bus stops,and the concentration point of morning and evening peaks happens to be in the opposite state.According to the spatial and temporal distribution characteristics of shared bicycles,hierarchical clustering and k-means clustering are used to divide the shared bicycle scheduling area;then the BP neural network and RBF neural network prediction models are used to predict the shared bicycle scheduling area demand.Both neural network training results show that the predicted value is close to the actual value,and the prediction error is controlled within 5%,and by comparison,the RBF model has higher prediction accuracy in such problems.The accuracy rate of more than 80% of the dispatch area is greater than 90%,and the highest accuracy rate is 97%,and the model fits well;Then,based on the prediction data of the shared bicycle scheduling area,reasonable assumptions are made,a regional shared bicycle scheduling model with time window as the goal of cost minimization is established,and the constraints of the model are proposed according to the objective reality constraints;Finally,a genetic algorithm is designed to solve the scheduling model,and the real data of Mobike in Nanjing is used as an example to solve a more reasonable scheduling scheme.更多还原