【作者】 贾晓秋；

【导师】 彭其渊；

【作者基本信息】 西南交通大学 ， 交通运输规划与管理， 2011， 博士

【摘要】 周期列车运行图问题一直是国内外研究的热点问题。本文在借鉴已有的周期与非周期列车运行图的研究成果基础上,探讨了我国客运专线周期列车运行图的规划条件、特点、模型以及算法等方面。着重从数学规划模型与计算机算法实现等层面研究了我国客运专线周期列车运行图的基本优化理论与方法,这些理论和方法主要有：1、在周期列车运行图研究方面,在充分研究国内外周期与非周期列车运行图现有的研究成果基础上,本文首先详细讨论了周期列车运行图的各类周期约束条件及规划目标,由此构建周期列车运行图的多目标混合整数规划模型,讨论模型中周期变换整数的范围,并在此基础上,将转化为以势差为决策变量的可行势差模型以及以周期势差为决策变量的圈周期规划模型,从数学上揭示了这三类模型之间的内在关系。为了设计算法方便,将周期列车运行图的多目标模型转化为具有多优先级结构的单目标模型来处理。2、在周期列车运行图的算法开发方面,本文提出了周期列车运行图与非周期列车运行图都适用的时空布线本质过程——列车-区间动态布线过程,并在此基础上,针对周期列车运行图的多目标模型,开发出基于列车一区间动态偏序的周期列车运行图的启发式算法；另一方面,同样基于运行线的时空布线本质过程,本文又根据列车运行图与Job-shop问题的相似性,提出了基于Job-shop的遗传算法。在用C++与Access2003开发的京津客运专线周期列车运行图的试验平台上,上述开发的两种算法均能够给出周期列车运行图不同速度模式下的满意解。3、在动车组运用计划的研究方面：在已有研究成果基础上,本文提出了非周期动车组运用计划的多目标规划模型,并将之转化为具有多优先级的单目标模型。针对非周期动车组运用计划的多目标模型开发了带检修回朔的遗传算法。并根据列车周期运行情况,本文将所建的非周期动车组运用计划的多目标规划模型转化为周期动车组运用计划模型,然后基于列车的周期运行性,提出了基于周期接续网络的检修回朔算法。因而可以在周期条件下大大简化动车组运用计划的编制问题。4、本文还简单地讨论了将非周期运行图的评价方法应用到周期列车运行图一些方法。周期列车运行图的规划系统较为复杂,本文采用类比与分析相结合,使用数学规划、周期事件规划、计算理论和算法分析等现代数学建模理论与方法进行相关研究工作。首先在对比分析国内外关于周期与非周期列车运行图建模思想的基础上,构建周期列车运行图的决策变量不同的三类模型。在周期列车运行图的算法上,通过构建列车运行图的时空布线的实际过程,开发出基于列车-区间动态偏序关系的周期列车运行图布线的启发式算法,运用类比的方法,开发出基于job-shop的周期列车运行图遗传算法。在动车组运用计划构建模型过程中,借鉴了现有的定性分析成果,构建非周期动车组运用计划模型,周期动车组运用计划模型。并结合这两大类动车组运用网络的各自特点有针对性地开发了带检修回朔的遗传算法与基于周期接续网络的检修回朔算法,通过对比,将非周期运行图的评价方法运用到周期运行图的评价当中。更多还原

【Abstract】 The problem of periodic train running map is one of the focus problems. On the basic of referring to the present theories and methods about periodic and non-periodic train timetable home and abroad, the dissertation discusses programming conditions, the characteristics, and models and algorithms on passenger dedicated lines in our country. It mainly researches on the basic optimization theories and methods of passenger dedicated lines in our country from the point of view of mathematic programming and algorithms implementation by computer. The theories and methods mainly have:1In the studies of periodic train timetable, on the basic of sufficiently discussing the present research fruits about periodic and non-periodic train timetable, the dissertation, firstly, researches the kinds of periodic constraint conditions and programming objective functions in the periodic train timetable problem, according to these conditions and objectives, secondly, constructs multi-objective mix integer programming model of periodic train timetable problem, discusses the taking value scopes of periodic changing integers, thirdly, transforms the model into feasible difference problem model whose making decision variables are differences between two any events and cyclic periodic programming whose making decision variables are periodic tension. The transforming uncovers the inherence relatives between three kinds of models from the point of the mathematics views. For the sake of conveniently designing algorithms, the multi-objective model is transformed into single objective model with multi-priorities structure.2In the developing algorithms of periodic train timetable problem, the dissertation proposes the natural procedure of the time-space assigning running lines, namely the procedure of train-section dynamically assigning running lines. By the base of the procedure, aiming at the multi-objective model of periodic train timetable problem, the dissertation designs a heuristic algorithm based on dynamically precedence orders of train-sections. On the other hand, by the same assigning running lines procedure, because of the similarity between job-shop problem and the periodic train timetable problem the dissertation develops a genetic algorithm based on job-shop problem. On the platform of Jingjin passenger train running map system by C++and Access2003, the above two algorithms can both give out satisfying solutions of periodic train timetable problem under the conditions of different train-speeds modes.3In the study of the train-set routing problem, on the basic of the present fruits, the dissertation proposes the multi-objective model about non-periodic train-set routing problem, and transforms it into the single one with multi-priorities structure. Aiming at the model, the dissertation develops a genetic algorithm with checking backtracking procedure. For the case of train periodically running, the dissertation rebuilds the multi-objective model about non-periodic train-set routing problem into the periodic model. Then the checking backtracking based on periodic connecting network is designed, which can simplify the train-set routing problem under the conditions of train periodically running.4The dissertation simply discusses the methods of applying the evaluation methods of the non-periodic train running map for the evaluation methods of periodic one.Since periodic train timetable problem is very complicated, the dissertation applies combination of analogy and analysis methods, uses the modern mathematics building model theories and methods, such as mathematics programming, periodic event scheduling programming, computing theories and algorithm analysis, to deal with the relative researches. Firstly, by the comparing and analyzing the building ideas of periodic and non-periodic train timetable home and abroad, the dissertation constructs the three kinds of models which have different making decision variables about periodic train timetable problem. On the algorithms of periodic train timetable models, by giving out the real time-space assigning running lines procedure about train timetable problem, the heuristic algorithm based on dynamic precedence orders of train-sections is developed. By the analogy method, the genetic algorithm based on job-shop problem is designed. In the process of building the model about train-set routing problem, the dissertation refers to the present qualitative analysis fruits, constructs the two models of periodic and non-periodic train-set routing problem. Considering the characteristics of the two kinds of models themselves, the dissertation targetedly develops the genetic algorithm with checking backtracking procedure and checking backtracking algorithm based on the periodic connecting network. Finally, by the contrast method, the evaluation methods of non-periodic train timetable problem are applied into the periodic one.更多还原