Car Flow Prediction And Dynamic Car Flow Organization Method In Railway Network

Railcar flow prediction is a problem of great concern to daily rail transport operation. To ensure the methodical transport order, it is necessary to predict the future flow in the network, grasp the possible changes of the flow, as well as make transport plans and adjust car flows in light of it. However, the existing method of railcar flow prediction is still limited to the rough calculation of transport time and the human judgment based on historical experience, which has brought out the obvious lag of the future information of car flow distribution in the network for the train operation department, and the unfavorable situation of marshalling yard/boundary congestion or train reservations at stations. In the present comprehensive transportation system, the railway transportation has become a bottleneck which restricts the economic development. Therefore, under the condition that the railway information construction is getting complete, the accurate real-time car flow prediction will make the railway transportation department grasp the changing trends of car flows in the network and then make rational and efficient response decisions to it, which is advantageous to guarantee the reasonable layout of car flows in the network and the optimal allocation of transport resources.Car flow prediction and dynamic car flow organization are complementary to each other. On the one hand, the conclusion of car flow prediction can provide the data parameters and decision support for the dynamic car flow organization plans. On the other hand, the simulation of car flow organization can help reach effective car flow prediction results. Therefore, car flow prediction must be studied combining with dynamic car flow organization. Only the combination and synthetic analysis of calculation and organization can promote the accuracy of car flow prediction and the rationality of daily transportation plan.As the concept of railway freight transportation is changing, the urgent needs to improve railway freight transportation service quality makes more and more experts and scholars begin to pay close attention to the problem of car flow prediction and dynamic car flow organization. However, the existing studies concentrated on the introduction of car flow prediction method and the study on static car flow organization such as rail blocking plan optimization. So there is a lack pf a set of systematic theories and approaches for car flow prediction and dynamic car flow organization. Therewith, based on domestic and foreign related researches, according to the characteristics of railway transportation organization in China, this paper detailed the nature of car flow prediction and dynamic car flow organization, and presented a series of targeted theories and approaches which have a positive meaning for improving the level of railway dispatching and enhance the competitiveness of the railroad.The major research contents in the dissertation are presented as follows:1. The basic principles of loaded and empty car flow prediction were addressed:In terms of loaded car flow prediction, the basic principles and methods of it were studied with taking the loaded flow’s transportation time as the breakthrough point. The difficulties how to calculate the transfer time of loaded car flows in classification yards along the running path were analyzed, and the theories that using rail blocking plan to determine the places where loaded car flows doing work with classification and using locomotive routing to determine the places where loaded car flows doing work without classification were given. In terms of empty car flow prediction, this paper forecast the spatial and temporal distribution of empty cars mainly by simulating the process of dynamic empty car distribution. The basic theory framework for dynamic empty car distribution was constructed by time-space networks. Besides, the advantages and disadvantages of discrete and continuous time-space networks applied in practical situation were discussed respectively.2. Considering the disadvantages that predicting the long-term (in 3-7 days) loaded car flow distribution in the network only depending on the running cars at the decision reference moment would result in the loss of many flows, in this paper, according to the time-space distribution mechanism of flow generation, operation and disappearance, the loading demand information was added into the parameters for long-term loaded car flow prediction, and we assumed that all the loading demands had been satisfied, then achieved the reasonable long-term potential loaded car flow prediction by the calculating the transportation time of the running cars and the required car flows. Referring to the results of the calculation, we can find out the rail line section of which the potential flow burthen exceeded its capacity, i.e. the bottleneck section. Thus, how to restrict the heavy burthen of the bottleneck section in order to ensure the smooth operation of car flows in the network is an important problem that needs to be solved urgently in railway dispatching, and it is also the extension problem of car flow prediction. In this regard, first, this paper gave the method of calculating the dynamic car flow structure of the bottleneck section in the target date based on the results of long-term loaded car flow prediction, and extracted the object for studying the dynamic allocation of the bottleneck section capacity accordingly. Secondly, we analyzed a number of indicators which affect the decision of allocating acknowledged cars, constructed a multi-objective programming model for the dynamic allocation of the bottleneck section capacity, and designed an algorithm to solve the model based on the weighted sum method. Finally, by taking Luoyangdong to Tielu section on Long-Hai Rail Line as an actual example of bottleneck section, we achieved the potential flow of the bottleneck section in the 5th day, gave the quota allocation of loading demands, and discussed the rationality of the scheme.3. For the problem how to calculate the empty car flow discharged over network boundaries which is extremely concerned about car flow prediction, in this paper, we proposed a solution to achieve the empty car flow distribution in the network by simulating the dynamic empty car distribution firstly, and then to track the empty car flow over each boundary. First, an isomorphic terminal network including 155 nodes was constructed instead of the whole network to make the calculation of dynamic empty car distribution feasible. Secondly, in order to make the problem more intuitive, a discrete time-space network was constructed by calculating the average runtime of empty car flow between terminals. According to its particularity, the optimization model for dynamic empty car distribution was constructed to provide a theoretical framework for solving the problem. Again, a chaos-based particle swarm optimization algorithm was presented based on the characteristics of the model and was applied to solve the dynamic empty car distribution problem at the terminal network level. Finally, in light of the computational conclusion of dynamic empty car distribution, the algorithm was given for calculating the next day’s empty car flow discharged over network boundaries4. For the micro-level car flow prediction, i.e. the calculation and prediction of loaded and empty car flow distribution at any moment in the near future (on the next day) and the blocking content of each outgoing train, taking any railway bureau network for the study, we simulated the daily operation procedure of loaded and empty car flows according to train schedule, and presented the scheme of car flow allocation in the trains. First, taking the train operation diagram as the basic framework, a continuous time-space network was constructed. The nodes and arcs in the network were designed according to freight car turnaround procedure. Secondly, in light of the operation characteristic of loaded and empty car flow in daily work, we used the path-based model thought to describe the features of dynamic loaded car flow and used arc-based model thought to depict the decision of dynamic empty car distribution, then constructed a model to optimize the organization of loaded and empty car flows in a railway bureau based on this modelling idea. Again, the genetic algorithm of integer coding was chosen to solve the model according to the analysis of calculation scale of real-life problems. Finally, taking the dynamic loaded and empty car flow organization in Lanzhou Railway Bureau for example, we effectively calculated the blocking content and car flow source of each freight train marshalled in this railway bureau. Meanwhile, this computational case indicated that the model and algorithm proposed in this paper was of great practical significance.