| During the Twelfth Five-Year Plan, China railway construction has a great development and built new railway line mileage of 30,500 Km. By the end of 2015, China railway operating mileage is 121,000 Km, and has formed a complex railway network. The single-track railway, the doubletrack railway, the three-track railway and the quadruple-track railway coexist, and intersect at the hub station. There are many block system coexist in the railway network, such as automatic block system, automatic inter-station block, semi-automatic block system and moving block system. Also there may be more than one train path between any two stations. The medium and high speed trains running in the same high-speed railway line and the high-speed railway line join the ordinary railway line in the hub station. All of these make the operating environment being very complex and hard traffic organization, which cause train delay easily. This article focuses on the train delay control Strategies and method for the complex railway network, give a in-depth analysis of the cause and the pattern of the train delay, propose delay-control models and methods based on static and dynamic traffic, and do some experimental verification.Firstly, The cause of train delay, the type of train delay, the mechanism of train propagation, the mechanism of delay absorption and the mechanism of delay prediction is detailed analysis. Noted that systemic factors are the main cause of train delay. The secondary delay, which caused by the primary delay and generates a domino-effict delay phenomenon, is the main cause of delay propagation. The train delay absorption is realized by using buffer time and remaining capacity of timetable. The basis of train delay prediction is knowing train position and disturbance distribution. In order to get the distribution of train delay, a heuristic funing-parameter distribution fitted method based on K-S test is presented and used for train arrival and departure delay distribution fitted. The results show Weibull distribution fitted best for the arrival and departure delay distribution, and the exponential Distribution is also fitted well. With the distribution of arrival delay distribution, a model base on Random Forests algorithm for predicting train arrival delay is proposed and the experiment results show that the accuracy of arrival delay has been great improved Compared with the prediction method besed on experience.From the perspective of allocating buffer time reasonablely, the criterions and indicators are proposed for determine the size of buffer time and allocating the buffer time. Then a buffer time allocation model based on continuous disturbance distribution and the analyticly solution method are presented. The theory experiment results show that the the more buffer time should be allocated to the first section in two section line. In order to simplify the computational complexity for the line with more sections, a discrete model is proposed with discrete disturbance distribution which is discrete based on the characteristic of train events. A Tabu Search method is used to get the optimal solution from the discrete model. Also the buffer time allocation in case of different sections and different match degree of buffer time and the disturbance are in detailed discussed.Finally, the delay control methods and strategies based on train operation adjustment are proposed. In order to elimination of potential delay and get more time for Scheduling decisions a train operation adjustment scheme on the basis of train confict prediction is presented. The a MIP model for train operation adjustment is built with the consideration of blocking type, signal constrains and buffer to improve the performance of train delay control. To genetic the MIP model for all blocking mode,an improvemen is made for the model. Then three algorithm, right shift, local search and iteration local search, are proposed to get the optimal solution from the improved MIP model and the experiment results are in detailed discussed... |
PDF Full Text Request