Zhongcan Li
School of Transportation and Logistics, Southwest Jiaotong University, Chengdu, China
Ping Huang
School of Transportation and Logistics, Southwest Jiaotong University, Chengdu, China / High-speed Railway Research Center, University of Waterloo, Waterloo, Canada
Chao Wen
School of Transportation and Logistics, Southwest Jiaotong University, Chengdu, China / High-speed Railway Research Center, University of Waterloo, Waterloo, Canada
Yixiong Tang
School of Transportation and Logistics, Southwest Jiaotong University, Chengdu, China
Ladda ner artikel![](/images/PDF_24.png)
Ingår i: RailNorrköping 2019. 8th International Conference on Railway Operations Modelling and Analysis (ICROMA), Norrköping, Sweden, June 17th – 20th, 2019
Linköping Electronic Conference Proceedings 69:45, s. 688-703
Publicerad: 2019-09-13
ISBN: 978-91-7929-992-7
ISSN: 1650-3686 (tryckt), 1650-3740 (online)
Primary delays (PDs) are the driving force of delay propagation. Hence, accurate predictions of the number of affected trains (NATs) and the total time of affected trains (TTATs) due to PDs can provide a theoretical background for the dispatch of trains in real time. Train operation data were obtained from Wuhan-Guangzhou High-Speed Railway (HSR) station from 2015 to 2016, and the NAT and TTAT influence factors were determined after analyzing the PD propagation mechanism. The NAT predictive model was established using eXtreme Gradient Boosting (XGBOOST) algorithm which was more efficient than other machine learning methods after comparison. Furthermore, the TTAT predictive model was established based on the NAT model using the support vector regression (SVR) algorithm. The results indicate that the XGBOOST algorithm has good performance on the NAT predictive model, whereas SVR is the best method for the TTAT model using Lessthan5 variable, which is the ratio of the difference between the sample size of actual and the predicted values in less than 5 min and the total sample size. In addition, 2018 data were used to evaluate the application of NAT and TTAT models over time. The results indicate that NAT and TTAT models have a good application over time.
High-speed railway, Primary delay, Number of affected trains, Total time of affected trains, Machine learning
Inga referenser tillgängliga