【作者】 徐杰；

【导师】 杜文；

【作者基本信息】 西南交通大学 ， 交通运输规划与管理， 2003， 博士

【摘要】 车站的调度指挥人员是通过车站作业计划来进行运输组织的，阶段计划是铁路车站作业计划之一。一个好的阶段计划应该反映出本阶段的工作重点，并按照列车编组计划的要求，把车流及时编成各种列车，保证本阶段的所有出发列车都能正点发车，并能合理地安排列车解编顺序以及利用到发线接发列车。目前我国区段站阶段计划的编制仍然停留在手工编制的层次上，编制速度慢，并且技术作业图表也主要是采用手工绘制，劳动量也高。阶段计划编制的自动化在现场中具有很大的实用价值，它将有效地提高阶段计划的编制速度及质量，进而提高路网的通过能力，并能很大程度上减轻车站调度员的工作负担。 要研究计算机自动编制阶段计划，必须分析铁路区段站的作业系统，通晓区段站作业系统的工作过程，找到区段站中作业组织之间的规律，才有可能编制出理想的作业计划。本文首先介绍了区段站必需的设备，及各种作业过程和全站的协调作业，为研究阶段计划的计算机自动编制打下基础。 应用计算机系统实现阶段计划的自动编制，在理论和实际应用上都很有意义。在实际应用中可以克服由于人的思维方式的不同以及专业素质的高低对计划编制质量的影响，提高车站的调度水平；在理论上可以进一步丰富和发展区段站作业计划编制的内容，推动数学理论向前发展。为了更好地说明下文所要解决的关键问题——建立合理的数学模型和寻找有效可行的算法，本文阐述了算法和启发式算法的理论基础，并扼要地介绍了遗传算法、模拟退火算法的主要思想和在解题过程中应用的重要参数，以及图的着色理论；及其在作业计划编制中应用的可能性。 对有关文献中建立的车流组织的网络模型进行分析，指出其不合理性。在现场调研的基础上对站调推算车流的思维活动进行全面的分析，找出其中可以借鉴的规律，得到对区段站应用计算机编制作业计划的必须的理论。依据上文提炼的理论，仿照现代经济社会中的交易活动，提出列车推算中的虚拟价格和收益函数，得到简单可行的车流推算模型。在车流推算的基础上，从应用计算机模拟区段站调度员思维的角度出发，通过分析车列占用调车机车时区集合的特点，使用划分时间片的方法建立合适的调车机车安排的图论模型，将调机运用问题转化为顶点具有加权的图的K-着色问题。将模拟退火算法和演化算法混合起来运用于该问题，设计有序的字符串编码方法，构造了基于罚函数的适应度函数，采用交叉和变异等技术，求解该模型。并提出了车列占用时区调整的相关算法。参照调车机车运用计划的求解，建立到发 西南交通大学博士研究主学位论文 第11 页线占用安排的数学模型，并将贪婪着色算法和演化算法混合起来运用于该问题，取得满意解。 在应用计算机实现算法中，提出了计算机编制阶段计划系统本质上是一个决策支持系统（DSS）。按照计算机实现的一般步骤，首先对该系统的结构功能进行分析：把区段站的作业计划编制过程看作是若干个组合优化问题的启发式搜索过程；为系统提供一个友好的人机交互界面；有关数据和规则在系统中的表达。然后规划系统的总体结构和系统各部分的功能。最后给出系统决策必须要经历几的个阶段与算法的实例和结果。更多还原

【Abstract】 The stage plan for railway depot station is one of the operative plans, by which the dispatchers in the station manage all transportation work. A high-quality operative plan should focus on the important tasks, organize car flow into all kind of trains according to the demands of the marshalling plan, ensure that all departure trains leave station on time, schedule exactly the decomposing and marshalling of all trains and suitably arrange all trains to occupy the arrival and departure lines. At present, railway depot station daily-shift operative plans are drawn up manually, which is inefficient, and the dispatchers draw the plan chart by hand, which is toilsome. Therefore, the realization of automatically making the daily-shift plan with computer is practical value on the spot, which is conducive to increase the speed of making plan and improve quality of operative plan, upgrade the capacity of railway and greatly lessen the dispatcher’s physical work.In order to study the problem of automatically making railway locomotive station operative plan and making a high quality plan, above all, we ought to analyze the operation system of the station, know thoroughly the procedures of operation system and make acquainted with regulations of all the work of the stations. Thus, the paper introduces all the essential facilities, the procedure of all operations, such as marshalling, decomposing, fetching and delivering train, of the stations and coordinate work of all department, so that the working of automatically making stage plan with computer is well grounded on.It is significant in theory and in effect that automatically making operative plan with computer. The spottiness of plans workout can be avoided because of people’s different thought patterns and professional characteristics, in effect; In theory, the method of making operative plan for district stations will be further developed and the relative mathematics theories will be boosted. To better elucidate the following critical problems -- building the rational mathematics model and searching the effective algorithms, this paper sets forth the basis of algorithms and heuristic algorithm. Meanwhile it compendiously introduces the main principles of genetic algorithm and the stimulated annealing algorithm, principal parameters during the course of solving problems and vertex coloringtheory.This paper analyzes the network model of cars flow organization related to my work, which was presented in the references and points out that it is improper. Analyzing in detail the mental activities, which the dispatchers have during the course of dispatching, I find out the rules used for reference and the essential theory meeting the demand of operative plans with computers aided. According to the regulations above, the paper presents the virtual price and profit function in the course of selecting cars for the departing trains and the model of the grouping cars into arrival trains. Based on the conclusion of organizing into trains and simulating the thoughts of dispatchers in district stations, this paper analyzes the features of time sets of shunting locomotives, constructs the reasonable graph model in the way of making out time sets, and converts it into the problem of graph coloring. The paper exploits then method of combining the simulated annealing algorithm with genetic algorithm for the question and designs the coding way which based on ordinal characters, builds up fitness function on the basis of penalty function, adopts the basal technology, such as crossover, mutation etc. to deal with the model and put forward adjustable algorithms for trains occupying time sets. Consulting on the method of arrangement of shunting locomotives, the mathematical model of the arrival and departure lines occupying arrangement is set up. I apply the hybrid algorithm that combines greedy-selector into genetic algorithm to the question and get satisfactory answers.During the process of algorithm realization with computer, this paper draws a conclusion, which the syste更多还原