Research On Railway Empty Container Distribution Based On Data-Driven

Container transportation plays an increasingly important role in railway freight transportation in recent years because of its high degree of standardization,convenient cargo handling,transportation safety and conducive to the development of multimodal transportation.However,due to the uneven distribution of China’s goods and population,the limitation of container ownership,the overdue use of containers by cargo owners and so on,the container flow is unbalanced in time and space,so the empty container reposition is inevitable.At this stage,the railway transportation departments complete the formulation of empty container distribution plan between different railway bureaus manually according to the general distribution of serviceable containers,which leads to the low level of refinement of empty container allocation.With the construction of railway information platform,large-scale container application data are generated,which provides new support for traffic organization optimization.Therefore,this paper studies how to use data-driven technology to effectively improve the prediction accuracy of empty container supply and demand and refine the empty container allocation process.The main work contents are as follows:(1)This paper introduces the development status and trend of railway container transportation in China,and combs the existing research results of empty container transportation and the application status of data-driven and short-term prediction technology.Combined with the application mode of data-driven technology,based on the historical application data of railway containers,this paper constructs a decisionmaking framework of empty container allocation based on data-driven,and specifically describes the data prediction process,prediction method selection in the data prediction module and modeling ideas in the empty container allocation decision-making module in this framework.(2)After analyzing the change characteristics of the historical empty container supply and demand data of the container handling stations and preprocessing them,the machine learning algorithms are used to construct the LSTM-SVR combined prediction model to realize the short-term prediction of the number of empty container supply and demand.The calculation results show that the prediction error of the combined model is reduced by about 20% compared with many other single prediction models,such as GM(1,1),BP,ARIMA,LSTM and SVR.(3)Combined with the prediction error,the predicted value of empty container supply and demand of each handling station is processed into triangular fuzzy number.Based on the existing research and aiming at minimizing the running distance of empty containers,the basic model with fuzzy constraints and hierarchical allocation optimization model of railway empty container allocation are constructed respectively,and the fuzzy constraints in the models are clearly transformed by using the fuzzy chance constrained theories.(4)A case is designed and the optimization solver CPLEX is used to solve the above models,which verifies the effectiveness of the empty container allocation optimization models and realizes the empty container allocation decision.At the same time,the calculation results show that the formulated empty container allocation schemes in the next stage are conducive to shorten the running distance of empty containers and realize the accurate allocation of empty containers for container handling stations.In addition,compared with the "node-to-node" empty container allocation scheme directly formulated by the basic model,in the two-stage scheme with the empty containers are allocated within a bureau first and then across the bureau formulated by the hierarchical model,although its empty container running cost increases,the number of empty containers allocated across bureau is reduced,and it is more feasible because it is conducive to the centralized and unified distribution of crossing bureau empty containers,release passage resources and improve the efficiency of traffic organization.In addition,in the clear conversion of fuzzy supply and demand,it is found that different confidence levels will directly affect the preparation of empty container allocation scheme.There are 39 figures,13 tables and 58 references in this paper.