城市轨道交通客流短期预测方法及实证研究

发布时间：2018-06-14 22:54

本文选题：轨道交通 + 客流预测　；参考：《北京交通大学》2012年硕士论文

【摘要】：随着城市规模的不断扩大、人口的不断增加、私人小汽车的逐渐普及,交通拥堵、环境污染、能源消耗等问题日趋严重,这与人们自由出行的愿望背道而驰,城市面临巨大的交通压力。准时、环保、舒适、运量大、效益好的城市轨道交通已经成为解决交通问题的主要发展方向。客流量是城市轨道交通路网规划与设计的基础,而前期的客流预测是城市轨道交通工程建设的关键技术。围绕城市轨道交通客流的短期预测问题,本文主要进行了如下的研究工作： 1、对北京市轨道交通清明节假期的客流数据进行统计和分析,得到了其时序特征、周期规律及主要影响因素。此外,对线路流量和进站量数据进行聚类分析,有效划分数据样本类型,为模型改进及案例分析提供可靠的数据支持。 2、针对RBF神经网络单一模型结构的局限性现状,构建基于时序特征的多模块加权神经网络预测模型,通过轨道交通线路流量的预测实例,验证改进后预测模型具有更高的预测精度。 3、针对支持向量机单一核函数的不足,构建改进的基于混合核函数的支持向量机预测模型,通过对进站量预测的实例分析,证明改进后支持向量机模型更好的拟合效果及在城市轨道交通短期客流预测领域的可行性。 4、考虑RBF神经网络和支持向量机两种单项预测方法在学习方法、建模方式及结构特征等方面的优势,将这两种预测方法有机的融合在一起,构建基于灰色关联度最大化的组合预测模型。通过清明节线路流量的预测实例,得到了组合预测方法比单项预测方法更大的关联度,证明组合预测模型优于单项预测模型。
[Abstract]:With the continuous expansion of the city scale, the increasing population, the gradual popularization of private cars, traffic congestion, environmental pollution, energy consumption and other problems are increasingly serious, which runs counter to people's desire to travel freely. The city is facing great traffic pressure. Punctuality, environmental protection, comfort, large volume and good efficiency of urban rail transit have become the main development direction of solving traffic problems. Passenger flow is the basis of urban rail transit network planning and design, and the prediction of early passenger flow is the key technology of urban rail transit engineering construction. Focusing on the short-term prediction of urban rail transit passenger flow, this paper mainly carries out the following research work: 1. The statistics and analysis of the passenger flow data of Ching Ming Festival holiday in Beijing rail transit are carried out, and the time series characteristics are obtained. Periodic law and main influencing factors. In addition, cluster analysis of line flow and incoming data is carried out, and data sample types are divided effectively, which provides reliable data support for model improvement and case analysis. 2. In view of the limitations of single model structure of RBF neural network, The prediction model of multi-module weighted neural network based on time series feature is constructed, and the forecasting example of rail transit line flow is given. It is verified that the improved prediction model has higher prediction accuracy. 3. Aiming at the deficiency of single kernel function of support vector machine, an improved support vector machine prediction model based on hybrid kernel function is constructed. It is proved that the improved support vector machine model has better fitting effect and the feasibility in the field of short-term passenger flow prediction of urban rail transit. 4. Two single forecasting methods, RBF neural network and support vector machine, are considered in this paper. Based on the advantages of modeling and structural features, the two prediction methods are combined together to construct a combined prediction model based on the maximization of grey correlation degree. Through the example of Ching Ming Festival line flow forecasting, the combined forecasting method is better than the single forecasting method, and it is proved that the combined forecasting model is superior to the single prediction model.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：U239.5;U293.13

【参考文献】