The Modle And Algorithm Of Wagon-flow Allocation In Dispatching System Of Marshalling Station Under Uncertainty

Marshalling yard is railway's important grass-roots units, mainly handling the break-up and the marshalling of freight trains, and the main function is to simply "production train", that is, all wagon-flow which will arrive at the station are classified, according to the freight train marshalling plan, train diagram and "technical regulation" relevant requirements of making up a variety of departure trains and in accordance with the provision time of train diagram or Daily Plan to punctual departure, and it also charges the regulation effect of wagon-flow and train-flow in coordination of point and line capacity in the railway network.Marshalling Intelligent Scheduling System is an important part of the Marshalling Integrated Automation System, and then the optimization of stage plan for marshalling yard is the key link and theoretical difficulties of the system. The core problem of stage plan is to decide the marshalling formation and sources of wagon-flow of departure trains that is the wagon-flow allocation problem. It is to arrange the train's break-up and marshalling scheme and allocation problem, making the work of marshalling smooth, orderly, improve transport efficiency and transport tasks,to ensure the smooth of the entire railway network. the wagon-flow allocation problem is divided into dynamic wagon-flow allocation and static wagon-flow allocation. Dynamic wagon-flow allocation determines the order of break up and marshalling and the sources of wagon-flow of departure trains by making in an allocation scheme of break up and marshalling. Static wagon-flow allocation gains an allocation scheme by deciding the marshalling formation and sources of wagon-flow of departure trains.Scheduling system in the marshalling yard, there are uncertainties, the uncertainty of existing information (such as arrival, break-up, marshalling and departure, etc.), but also job uncertainty (such as train to, solution, code, send jobs) There's uncertainty conversion (such as the number of cargos the train compiled.) For the phase, the short span of time, information that is certain, the uncertainty is mainly reflected in the operating time and number of cargos on the train compiled. In the four operations, that is arrival, break-up, marshalling and departure, Arrival and departure operation time are relatively stable,but break-up and marshalling operation time are greater uncertainty, in which change of break-up operation time is larger. To more objectively and accurately estimate the train operation time, learning from the staff's scheduling experience, analyzing the various uncertain factors of the train operation time and considering the equipment and work stations having different organizational methods. In this papers, basing on probability theory, the use of regression analysis and parameter estimation to a reasonable estimate of operating time.Providing time of break up and marshalling operation are set as the fuzzy variable according to fuzzy mathematics, under certain confidence level operation time are expressed by pessimistic value of variables, and maximum number of vehicles in a stage plan are set as the target, the model of dynamic wagon-flow allocating under the uncertain condition can be built. And uses Approximate Nondeterministic Tree Search (ANTS) for solution dynamic wagon-flow allocation in the uncertain train operation time. The method is ACO that is one of Mathematical Program, in which time constraints is guaranteed based on defining the possibility break up set, waiting set and choice set to mapping dynamic wagon-flow allocating onto scheme tree. Improved route construction rule and pheromone updating, and the judgment of constraint condition in the process of transfer can enhance the efficiency of solution and convergence speed of the algorithm.In static wagon-flow allocation, the main uncertainly is the volatility of the number of cargos compiled. The problem of static wagon-flow allocation can be transformed into the problem of balanced transportation of fixed cost by constructing the network model of static wagon-flow allocation. By setting the virtual arriving train and dividing the departure trains into compulsive full axis and optional full axis, the objective function can be transformed into working out the minimum number of virtual arriving train's wagons. Model solved by the use of neural network algorithm. The algorithm gives the initial value to virtual arriving trains firstly, apply the learning rules and contact number of cargos compiled, tonnage rating and equated length of train to guarantee the full axis of departure trains in the process of calculating, and finally work out the minimum virtual arriving trains which is meant to realize the allocation scheme. Optimal solution obtained through the step by step iterative, with the ultimate complete solution to flow problem.