Various modeling exercises are presented to better interpret the function of urban rail transit systems.
An Improved Framework for Origin-Destination Prediction in Urban Rail Transit Based on Stack LSTM
BAO WANG, Southwest Jiaotong UniversityShow Abstract
Ailin He, Southwest Jiaotong University
luo xia, Southwest Jiaotong University
YING ZHU, Southwest Jiaotong University
Generally, the accuracy of passenger flow prediction in urban rail transit (URT) network is low due to the notable differences among different Origin-Destination (O-D) pairs. To improve the prediction performance of O-D flows in URT network, an improved framework which combines k-means clustering and stack long short-time memory (LSTM) is proposed (CS-LSTM). Firstly, the Silhouette Coefficient is calculated based on four indices, which is applied to determine 4 categories of O-D pairs. Secondly, in order to reduce the negative influence of random factors, the input data structure is adjusted by setting dynamic minimum thresholds of passenger flow. Thirdly, the parallel stack LSTM network is trained by utilizing the advantage of LSTM in predicting variable-length sequence data, and the two output sequences are re-constructed as ultimate outcome. Finally, a case study is conducted by using the data from Chengdu URT, where the proposed method is verified for different categories of O-D pairs, and the prediction error are compared between the proposed method and other 7 state-of-the-art methods. The result indicates that the proposed method has better prediction performance, especially in capturing the trends of passenger flow. We also compare the prediction performance under different time granularities with 15min, 30min, 60min interval. The results show that when the time granularity increases, the prediction error of O-D pairs with noticeable morning and evening peaks decrease significantly, and the prediction accuracy could be further improved through the fusion of different time granular data.
Short Term Passenger Flow Forecasting for Sparse OD: A Feature Simplification Based Machine Learning Approach
Keyang SUN, Southwest Jiaotong UniversityShow Abstract
Di WEN, Southwest Jiaotong University
Yan ZHANG, Southwest Jiaotong University
Hongxia LV, Southwest Jiaotong University
Short term prediction of OD passenger flow is essential for arranging network level proactive train rescheduling measures and passenger guiding measures. Many existing methods have obtained satisfactory predicting accuracy from the aspect of average predicting errors of all OD pairs. Our preliminary study showed that vast majority of OD pairs have small passenger flow or even zero passenger flow at small intervals. These OD pairs (defined sparse OD pairs) obtained larger predicting errors than average predicting error, but the predictability of sparse OD is proved to be quite high. Therefore, we propose a feature engineering based lightGBM (FE-lightGBM) for predicting sparse OD passenger flow in metro. The feature engineering module extracts several sorted OD features having strongest correlation with OD pairs from numerous OD passenger flow time series. The prediction part concatenates the extracted OD features with timing features and analogical features as the input of lightGBM. We also take Chengdu Metro as an example to testify the performance of feature engineering and FE-lightGBM by comparison with XGB, GRU/FE-GRU and SVR/FE-SVR. The results show that FE-lightGBM has high predicting accuracy and smallest calculating time; feature engineering has limited effect on different models due to the heterogeneity of each model’s input data structure. This study provides insight for improving sparse OD passenger flow’s predicting accuracy and a hint to predict classified OD pair passenger flow individually.
Optimal Utilization of Regenerative Energy by Using Power Profile for Urban Rail Transit
Juanjuan Cai, Beijing Jiaotong UniversityShow Abstract
Jing Xun, State Key Laboratory of Rail Traffic Control and Safety
Xiangyu Ji, Academic of Railway Sciences(Beijing) Engineering Consult Co.Ltd
Lei Yu, Communication and signaling branch of Beijing Metro operation Co. Ltd
Urban rail transit (URT) has developed rapidly in modern cities, its energy efficiency is getting more and more attention. The utilization of regenerative energy (URE) is an important method for energy-efficient operation of URT. Regenerative braking is an energy recovery mechanism that slows down a moving vehicle by converting its kinetic energy into electric energy. Other trains could use these regenerative energy by accelerating in a cooperative way. In order to maximize the URE, this paper puts forward an energy calculation method which considering regenerative braking power to optimize the timetable. First, we define 4 operating Modes of energy utilization, and formulate an integer programming model. Second, we design a branch and bound algorithm to solve the optimal timetable at different scenarios. Third, based on the operation data from the Yanfang Line, Beijing Metro, we evaluate the model. For peak hours, the results illustrate that compared with the original timetable, the proposed method can significantly improve the URE by 73.7%. For off-peak hours, it can improve the URE by 46.3%. Finally, the comparison between the proposed method and the kinetic energy based method is given. The proposed method could increase the URE by 29.75% and 9.93% for peak and off-peak hours scenarios respectively.
Urban Rail Transit Efficient Path Search using Best-First Strategy
Zhan-ru Liu, Southwest Jiaotong UniversityShow Abstract
Yi-chen Sun (email@example.com), Southwest Jiaotong University
Jie Zhang, Southwest Jiaotong University
Wen-cheng Huang, Southwest Jiaotong University
Min-hao Xu, Southwest Jiaotong University
The passenger route selection and passenger assignment are the basis of the analysis of the Urban Rail Transit (URT) passenger flow. A fast and efficient path search algorithm is of great significance for improving URT passenger management and operation planning. This paper reconstructs the urban rail transit network based on the network topology and proposes an efficient path search algorithm for urban rail transit based on the Best-First search strategy, which can improve the efficiency of the search while avoiding the problem of omitted or redundant routes. Firstly, reasonable restrictions were made on the efficient paths between ODs based on travel costs and transfer counts; secondly, the network structure was simplified by omitting intermediate stations; finally, a tree-like search based on the Best-First search strategy was performed on the simplified network to obtain efficient paths that meet the actual passenger choice. Based on the assumption that the travel time distribution of a single path approximately obeys a normal distribution, Origin-Destination (OD) pairs of Chengdu Metro were taken as examples to examine the algorithm. Travel time of specific OD pairs was extracted from the transaction data of Chengdu Metro and was analyzed by the Expectation Maximum (EM) clustering method to obtain the observed travel time of each route. The error between the clustering result and the efficient paths cost obtained by this algorithm was less than 5%, which verified the effectiveness of the effective path generation algorithm.
How to define the optimal catchment area for subway stations? A study of five cities in the U.S.
Hongtai Yang, Southwest Jiaotong UniversityShow Abstract
Xuan Li (firstname.lastname@example.org), Southwest Jiaotong University
Chaojing Li, Southwest Jiaotong University
Qian Ge, Southwest Jiaotong University
Yugang Liu, Southwest Jiaotong University
Direct demand modeling is a useful tool to estimate the demand of a subway station and to determine factors that significantly influence the demand. The construction of the direct demand model involves the determination of the catchment area. Although 800 meters are usually used, whether this distance could well represent the catchment area is not well studied. This paper investigates the subway station’s optimal catchment areas using data of five cities in America. We define five catchment areas by drawing buffers around each station with radius ranging from 300 to 1500 meters with the interval of 300 meters. Built environment characteristic within each catchment area are extracted as explanatory variables. Annual average weekday ridership of each station is used as the response variable. Three commonly used regression models, linear regression, log-linear regression and negative binomial regressions, are applied to examine which catchment area has the highest goodness-of-fit. The effect of distance decay is also considered by applying lower weight to outer catchment area. We find that the results of different catchment areas are comparable. Considering the effect of distance decay could barely improve the model results. When the catchment areas are overlapping, dividing the overlapping area by the number of times of overlapping could improve model results. The goodness-of-fit of the three models are comparable, though the log-linear regression has the highest prediction accuracy. This study could provide useful reference to researchers and planners on how to select catchment area when constructing direct demand models for subway stations.
DISCLAIMER: All information shared in the TRB Annual Meeting Online Program is subject to change without notice. Changes, if necessary, will be updated in the Online Program and this page is the final authority on schedule information.