The Research On Intelligent Job-execution Of Marshalling Station Synthetic Automation System

With the completion of the backbone of high-speed railway, railway passenger transportation is gradually transferred to the high-speed railway, causing the change of the structure of point-line capacity of railway freight. The ability of the "point", such as marshalling station, district station, is becoming a limiting factor. Under the background, marshalling station should meet the needs of the railway productivity layout adjustment, constantly adjusting and upgrading, greatly improve the operation efficiency and operation level, so arises at the historic moment a new generation of marshalling station integrated automation technology, which called SAM. In many technologies of SAM, integration and centralization are the foundation; information reflects the communication performance; automation embodies the control ability; intellectualization at the kernel, reflects and determines the overall level of the system.From a macro perspective, the intellectualization of marshalling station can be divided into three major areas: intelligent operation decision, intelligent scheduling and execution, intelligent equipment control. The intellectualization of the job execution of marshalling station belongs to the second area. The focus of the study is, on the premise of that the wagon-flow allocation and the sequence of break-up and make-up are determined, the intelligent allocation equipment resources which operation needed.In the dissertation, the marshalling station transportation business is analyzed firstly, including the role and status of marshalling station in current railway transportation planning and dispatching system, the main work and the overall process, three-level plan system, the cooperation and the relationship between multiple posts during the execution of the work. It is pointed out that, due to the enormity of drawing stage plan, using the stage plan to dispatch and command field work has not been truly implemented in many marshalling stations. Most of the stations often take the stage plan as a leading role and combined with the approach of "hierarchical processing, divide and rule, multi posts collaboration" to handle the scheduling and the execution. In the whole process, station dispatcher as the leader, with shunting plan maker, shunting warden, field watchman, promptly exchanges information, to achieve collaborative decision-making and cooperative job-execution.Then, the mathematical model of job-execution system is proposed, which simplified the multi-objective optimization problem as a bi-objective programming model with the optimization goal of the minimum execution cost on the basis of the maximum number of departing on-time trains. At the same time, this paper analyzes the constraints of the model, and to meet the needs of the hierarchical management and the hierarchical decision, factors affecting the train departure time be separated into two parts, namely fixed basic work time and dynamic operation delay time. The first part is determined by the station dispatcher, within his scheduling and sorting-formatting decisions; the second part is further decomposed into several parts caused by the arrangement of resources such as arrival-departure lines, sorting lines, shunting locomotives, humps, routes, which respectively determined by shunting plan maker, shunting warden, field watchman, within their resource allocation scheme.Based on the business analysis and the mathematical modeling, the requirements of the intellectualization of job-execution were analyzed, and the detailed description of the design solutions as well as the technological framework. It is needed to solve the optimization problem that absorbing the thought of large scale system theory of "reduced order model, hierarchical structure, autonomous decentralization, man-machine coordination" and constructing integrated solutions from multiple angles and multiple levels.The reduced order model referenced the singular perturbation method: firstly the reduction of the model’s order, each separated variable analyzed and calculated, and then the calculation results brought together for coordination and re-optimization in the overall level.The hierarchical structure is the main structure of large scale systems and has many advantages such as simplifying the analysis and synthesis of the system, reducing the difficulty of calculation and the amount of calculation, saving the calculation time, convenient management and control, high reliability etc. Corresponding with the dispatching organization structure, the marshalling station intelligent job-execution system adopts three-level hierarchical structure composed of scheduling level, organization level, control level.The combination of distributed artificial intelligence and autonomous decentralized system formed intelligent autonomous decentralized system. The intelligent autonomous decentralized system of marshalling station is divided into three groups: system management agent group, backstage computing and service agent group, interactive agent group. Each intelligent body in the system is an autonomous unit, communication between them only through the agent data field to implement to ensure their autonomy and weak coupling.The combination of human intelligence and machine intelligence is one of the most effective methods to solve the optimization problems of large scale systems. The construction of human-machine coordination system needs to grasp two things: the reasonable human-machine division principle and the friendly man-machine interface.This paper proposes four-layer technological framework, composed of four-base layer, intelligent support layer, back-end service layer, front-end layer. In the intelligent technology, taking into account that the job-execution involves many complicated and uncertain factors, a hybrid intelligent system using a variety of intelligent algorithms and techniques is suggested.Finally, it is studied that the intelligent arrangement to arrival-departure lines, sorting lines, humps, shunting locomotives and routes, including mathematical modeling, algorithm designing and example analysis.In the intelligent arrangement of arrival-departure lines, a model is proposed with a goal of on the basis of the shortest wait time considering the principle of flexible-use. Genetic algorithm is chosen for optimization.In the intelligent arrangement of sorting lines, we brought forward a model, with an object function of on the basis of the shortest dynamic time considering the principle of “regular use and flexible borrow”, and expert knowledge reasoning combined with genetic algorithm as the optimization algorithm.In the intelligent arrangement of humps a model is put forward, which targeted the shortest wait time, and chosen production expert system as its optimization algorithm.In the intelligent arrangement of shunting locomotives, we present a model with an optimization objective, which on the basis of the shortest dynamic time considering locomotive’s fixed work area, travel distance and the degree of the interference to other workings. BP neural network technology is used for optimization.As to the intelligent arrangement of routes, we take the shortest time with the combination of the factors of work types, work priority, degree of urgency, the distance as the optimization objective of the model, expert knowledge reasoning with ant system for optimization.