| There are two key technologies in synthetic automation system of marshalling yard, e.g. optimal establishing and dynamic adjusting the operations plans. In order to reduce the average dwell time of railcars and improve the efficiency of the railway freight transportation, operations plans should be elegantly designed to schedule the operations at the marshalling stations. In the three-layer dispatching system of the marshalling station, stage plan is the core of the operations plans, which is not only the specific working arrangements of each shift plan, but also the main basises of the shunting operation plans, which are critical to be used to schedule the operations and assign the resources for about 3 or 4 hours.At present, researches on optimization of stage plan at marshalling station are rich, but only some sub-problems i.e. determining the orders of the disassembly and assembly operations and the wagon-flow allocation are solved independently in the former researches, and the stage plan was unable to be optimized globally. At the same time, the complex operating environments were not considered and the domain reduction techniques were not used in the former researches, hence, the models were not adaptable and the algorithms were not efficient. Furthermore, the influences from the exception events were not taken into account in the most former researches, so that the stage plans were lack of robustness. The above defects made that the existing theoretical achievements about the optimization of the stage plan were failed to put into practical application. Consequently, the stage plans were established by the station scheduler manually using computer-assisted technology at present. In order to maximize the efficiency of the station and stability of the stage plan, the following key contents of the optimization of stage plan are studied in this dissertation:1. Based on the theory of Stochastic Petri net modeling and analysis, the performances of the marshalling station operation processes were analyzed qualitatively and quantitatively. Firstly, the stochastic Petri net (SPN) models of the operation processes of the marshalling station are built to analyze the dynamic characteristics and conflict relations qualitatively. On the basis of the qualitative analysis, the resource shared stochastic Petri net (RSSPN) models are set up. In the RSSPN models, influences of resources on the jobs are fully considered, the methods to calculate parameters and the technologies to simplify the models are proposed. Finally, the relationships between infrastructure configuration and processes performance are calculated quantitatively through the simulation tools of time Petri nets and MATLAB.2. On the basis of the analysis of the marshalling station operation system, the optimization problem of the dynamic wagon-flow allocation is solved using constraint-based cumulative scheduling, lexicographic multi-objective optimization, integer programming and multi-objective two-stage optimization theory. Firstly, a lexicographic multi-objective constraint optimization model of preliminary wagon-flow allocation is expressed as constraint predicates for the purpose to maximize the total number of the departure trains’priorities, minimize the average dwell time of the railcars and maximize the resource utilization. The model is solved iteratively through a hybrid algorithm of constraint propagation and multi-point constructive search. On this basis, an integer programming model of secondary wagon-flow allocation is built to minimize the number of the inbound trains which provide railcars to each outbound train. Then, Greedy algorithm is designed to optimize the initial wagon-flow allocation plan. This hierarchical approach improves the efficiency of the shunting operations and the redemption rates of the wagon-flow allocation plan.3. Considering the optimization of the dynamic wagon-flow allocation as the core, the job scheduling and the resources allocation problems are solved using constraint-based cumulative scheduling, integer programming and lexicographic multi-objective multi-stage optimization. Firstly, the constraint optimization models of transit trains and remaining inbound-disassembly trains are set up, which are solved by the hybrid algorithm of constraint propagation and multi-point constructive search. On this basis, the integer programming models of resources allocation are established, which are solved to allocate resource to each operation by Greedy algorithm. Finally, the multi-objective and multi-stage optimization model of the traffic-concentrating-based-on-directions stage plan is built, which realizes synthetic optimization of each sub problem.4. When the disassembly and assembly operations are determined, an integer programming model of flexible utilization of shunting tracks is set up, which takes into comprehensive consideration the changes of the cars concentration states, the marshalling yard capacity limit, the precedence relationships of the disassembly and assembly operations, etc. The inbound-disassembly trains are divided into some wagon-groups by the "cutting" algorithm and the shunting tracks where the wagon-groups will concentrate are variables in the model, and a heuristic backtracking algorithm with the value ordering heuristics technology is designed. Finally, the multi-objective and multi-stage optimization model of the traffic-concentrating-based-on-shunting-tracks stage plan is built, which realizes synthetic optimization of each sub problem.5. Based on the optimization of the static stage plan, the robust dynamic adjustment problem of the stage plan is solved through minimal perturbation in dynamic cumulative scheduling theory. Firstly, the robust dymanic adjustment model of the stage plan is established. In the adjustment model, maximizing the robustness of the stage plan is considered as the primary objective and maximizing the benefits of the marshalling station is taken as the secondary objective. Then, the heuristic rules to revise the model dynamically with the exception events are constructed. Finally, the robust model is solved by the hybrid algorithm of constraint propagation and improved backtracking.6. In order to make the theoretical achievements of the optimization of the stage plan can be applied, the optimization system for stage plan is designed and achieved through the object-oriented programming technologies. Finally, the system has been embedded in the synthetic automation system of marshalling station.This dissertation studies some key issues of the optimization of the stage plan, e.g. the analysis of marshalling station operation system, the optimization of dynamic wagon-flow allocation, the synthetic optimization of stage plans and the robustness of dynamic adjustment of stage plans. And great theoretical achievements has been achieved, and the optimization system of stage plan at marshalling station has been relized use computer software technologies. Instance validation results indicate that this system has high adaptability, reliability and high efficiency, which can can do some aided scheduling for the station dispatcher automatically and intelligently. |
