Study On Dynamic Service Network Design With Resource Management In Rail Container Transportation

In recent years,railway freight has developed rapidly for its low-pollution and efficiency.Firstly,the government has promoted the transportation structure adjustment plan of "Transferring the freight from road to rail" in 2018,and the State Railway Group has also formulated the action plan of freight increment.Secondly,the expansion of the railway network and the development of high-speed railway passenger transport have greatly released the transport capacity of railway freight.In the foreseeable future,the railway freight organization will be more refined to ensure the delivery period of goods and improve the customer service level.Based on a new mode of railway container transportation and using the operational research method,the dynamic service network with resource(e.g.,containers,train stock,etc.)management is studied.Specifically,the freight flow transportation organization and transportation resources(such as containers,train stocks,etc.)are integrated into a time-space network.Resource management refers to two kinds of constraints.One is the “design-balance” constraints of train stocks(the number of train stocks originating and arriving at each terminal is equal),and the other is the “circle-route” constraints of train stocks(the train stocks start from the originating terminal where they ultimately ought to return).Resource management is used to ensure that there are enough resources at each terminal in the next planning horizon.A model is built based on a time-space network and solved to obtain the schedule of train stocks and the routing of shipments through the service network within the planning horizon,which provides a complete plan for freight transportation and then ensures the delivery due dates of shipments.The detailed contents of the research include:1.Based on a cross-border rail network,the aim is to guarantee the transit period of shipments and minimize the operation cost.Firstly,a periodic discrete space-time network is constructed based on the given Hub-and-Spoke network to represent the fleet management and shipment transportation,and all possible space-time paths of each shipment are obtained by enumeration.Then,considering the constraints of fleet circulation and container circulation,and the limited number of space-time paths chosen by each shipment,etc.,a mixed-integer model was built,which integrates cargo flow,container flow,and train flow.A physical network of China Railway Express with 8stations and 7 sections was established for the numerical experiment.Five test cases with different solving scales were designed and solved by the commercial software,i.e.GUROBI.The satisfactory solutions can be obtained within 1 hour in all test cases,which demonstrated the validity of the model.2.Because of the complexity of DSNDRM,a two-stage method is proposed to decompose the DSNDRM problem into two subproblems,i.e.,dynamic service network design problem(DSND)and train resource allocation problem(TRAP).The DSND focuses on selecting and scheduling services without considering the usages of train stocks,while the TRAP concentrates on the assignment of train stocks.For DSND,shipments are required to be delivered by exactly one train.By enumerating all possible consolidates of shipments,an operational network is built,so as to allow a joint optimization of the flow-to-service assignment,which tightens the lower bound of the model and improves the solving efficiency.Both the arc-and path-based models for DSND are given.The former one can be solved by GUROBI,and the latter one is solved by a bespoken Branch-and-Price algorithm.For TRAP,a “transportation problem” is solved to introduce deadheading services to ensure the “design-balance” constraints can be satisfied.Two types of TRAP models are given,that is,TRAP models without shifting and with shifting the result of DSND.The numerical experiment is based on the railway network in China.The result shows that the two-stage method is effective and efficient compared with the integrated model.3.The DSNDRM with “circle-route” constraints are contemplated,which is also decomposed into two subproblems,i.e.,DSND and TRAP.Shipments are not required to be delivered by exactly one train stock.Thus,the DSND is to determine which transportation arcs are needed to be selected,rather than selecting and scheduling services.Since the arc-based model for TRAP is difficult to solve,it's reformulated by introducing path variables for the “circle-route” constraints.A solution method that combines Benders decomposition and column generation is proposed to solve the path-based model.To improve the solving efficiency of column generation,a primal-dual column generation method(PDCGM)is applied to improve the solving efficiency of column generation.The numerical results show that the proposed algorithm is more than GUROBI to solve the large-scale instances,and PDCGM can improve the efficiency of column generation.4.Taking China Railway Express as an example and using the two-stage solution method in Chapter 5,three different operation rules of train stocks are compared,that is,without the balance-constraint,with the balance-constraint,and with the circle-constraint.Compared with the former one,the latter two ensure that there are enough train resources to complete the schedule plan in the next stage.Meanwhile,the efficiency of the algorithm with and without the PDCGM acceleration strategy is compared,and the effectiveness of the algorithm is verified.