A Study On The Multi-Classification-Yard Location And Improvement Problem Based On Shipment-Block Sequence

Classification yards are usually referred as hubs for freight train formation,reclassifying a great many of railcars.In recent years,with the expansion and improvement of railway network,some classification yards have connected to more rail lines,whose importance have been enhanced.As the percentage of double-track railway and electrified railway increases,the problem of mismatch between yards and rail lines becomes more acute.Due to the development of high-speed railways,many existing rail lines are capable of handling more freight trains.In the adjustment of transportation structure,some freight flows shift from high way to railway,which results in the growth of railway freight volume and classification yards workloads.Under the background,improving(or building)some yards is of great significance.Generally speaking,a modern classification yard costs hundreds of millions of dollars and occupies a land of several square kilometers.To avoid a huge waste of resources resulted from unreasonable establishment or improvement of yards,it is necessary to study in-depth the classification yard location and improvement problem.Since the workloads of yards are interrelated,i.e.,the change of reclassification location of a certain shipment might impact the workloads of relevant yards,the yard location and improvement problem not only refers to the selection of investment plans,but also the determination of shipment-block sequence(i.e.,train formation plan).However,train formation plans are developed mainly relied on manual process without any optimization-based approach.And the decision-making of yard location is basically based on expert experience.Moreover,the complexity of yard location and improvement problem grows exponentially with the number of candidate yards and the scale of railway network,hence it is increasingly difficult to solve yard location and improvement problem manually.In this case,adopting computer-aided decision-making tools is more acceptable and feasible.There are a fair share of studies for the joint optimization of train formation plan and car flow path,however,a few research literature is devoted to the joint optimization of classification yard location and train formation plan.Similarly,multi-period location problem and hub location problem are well-studied,while multi-period classification yard location and improvement problem has not attracted any attention in the operations research literature.Furthermore,there are a lot of differences between China and North America in the rail operation plan,which makes it difficult to directly adopt existing research results.In order to provide a solid aid for railway operation and management officers,as well as survey and design practitioners,this dissertation mainly focuses on the problem of ideal workloads of classification yards,single-period classification yard location and improvement problem,and multi-period classification yard location and improvement problem.The research contents in detail are described as follows:(1)The main function of classification yards and major factors influencing the location of yard are analyzed.First,we introduce the operation contents of yards,and describe the differences between classification yards and district stations.Second,we analyze in depth the workloads and investment costs of yards,as well as the railway network structure,which are three major factors.Then,the impact of urban development,engineering geological condition and high-value commodity on yard location is analyzed.(2)The problem of ideal workloads of classification yards is studied.The car flow path and train formation plan are jointly optimized,withought considering the constraints of classification capacity and shunting tracks,aiming to obtain the ideal workloads of yards.A non-linear 0-1 programming model is constructed,whose objective is minimizing daily operation cost,i.e.,the sum of railcar accumulation cost,reclassification cost and transportation cost.Given the difficulty in solving the model which results from non-linear terms,linearization techniques are adopted.A small-scale numerical example is solved by using commercial software Gurobi and simulated annealing algorithm respectively,in order to test the efficiency and effectiveness of our model and approach.(3)A single-period classification yard location and improvement problem is investigated.First,major types of classification yard are analyzed.Then,binary variables are introduced to describe key factors in yard investment and train formation,and a bi-level programming model is proposed.The upper-level program is intended to find an optimal establishment or improvement strategy for candidate nodes,considering budget constraint,logical constraints,etc.The capital recovery factor is introduced to annualize investment,and zero investment is treated as a special strategy.The objective of the upper-level program is minimizing the sum of annualized investment and annual operation cost.The lower-level aims to obtain the least costly plan of railcar reclassification on the basis of the strategy given by the upper-level.The objective of the lower-level program is minimizing the daily operation cost,considering the constraints of yard reclassification capacity,shunting tracks and segment capacity,as well as logical constraints between decision variables.In practice,the daily workload of a yard might fluctuate,hence,a proportional factor of the original capacity is introduced to avoid overloaded conditions in peak hours.Similarly,linearization techniques are employed to transform our model into a linear one.Moreover,a small-scale instance is also carried out.(4)A multi-period classification yard location and improvement problem is addressed.The whole planning horizon is divided into several sub-periods,and the staged investment problem of yards is solved.A bi-level programming model is established.The upper-level program aims at minimizing the sum of capital investment and operation cost throughout the whole planning horizon,while the lower-level program is intended to minimize the sum of net present value of daily operation cost in each period.Likewise,a small-scale numerical example is solved to verify the effectiveness of our model and approach.(5)A real-world case study is carried out based on China railway network.As Gurobi cannot solve large-scale real-world yard location and improvement problems,a simulated annealing algorithm is adopted.A condensed network is constructed on the basis of China railway network,and the input data is pre-processed.Then,a real-world instance is solved by using simulated annealing algorithm,and sensitivity analysis is conducted.