基于协同指挥的高铁智能调度理论与方法研究

The intelligent train dispatching is a critical point of intelligent high-speed railway （HSR） operation. Multi-profession coordination of dynamic dispatching, cooperative decision-making, and emergency management is a significant issue that requires to be urgently solved. This project starts with the analyzing of the dispatching essence and its entire process, regarding the assessing, developing and protecting of transportation states as the clue. Based on large-scale and complex train operation data, various advanced methods, such as data science, artificial intelligence, operation research and synergetic, are synthetically used to study the closed-loop control problem of intelligent traffic control, which is defined as from train state assessment and deduction, train rescheduling, collaborative dispatching, and emergency management to train operating state protection. Then, the coupling mechanism among train operating environment, real-time risks and train operation states is investigated. Besides, the train operating state assessment and developing theories based on the multi-driven of time and events are proposed. Also, train delay propagation and recovery mechanism are revealed; data-driven train delay recovery and intelligent train rescheduling methods in various scenarios are established; and knowledge automation of intelligent train rescheduling is constructed. Additionally, cooperative decision-making by multi-dispatchers for HSR dispatching is proposed, and a train state protection method from point and line to a local railway network is introduced. Finally, the prospected results can provide theoretical and technical support to the intelligent train dispatching and subsequently the construction of HSR.

智能调度是智能高铁建设的重要内容，多工种协同的动态调度、协同决策、应急处置等方面的基础理论研究是智能调度亟待解决的难题。本项目从剖析高铁调度指挥本质和全过程着手，以运输态势评估、演化与保障为研究脉络，基于调度指挥多源信息，综合运用数据科学、人工智能、运筹学、协同论等理论和方法，研究高铁智能调度闭环控制“运输态势评估及推演-列车运行调整-多工种协同指挥及应急处置-态势保障”的基础理论。探明高速列车运行环境、实时风险与列车运行态势的耦合作用机理，提出时间-事件混合驱动的列车运行态势评估及演化理论；揭示高速列车晚点传播-恢复机理，建立数据驱动的列车晚点恢复理论以及不同情景下的列车运行调整智能化理论，构建列车运行调整自动化知识；提出高铁调度指挥多工种协同决策机制，建立面向“点→线→面”态势保障的高铁调度协同决策智能化理论与方法。研究成果可为实现我国高铁智能调度、加快智能高铁建设提供理论和方法支撑。

智能调度是智能高铁建设的重要内容，多工种协同的动态调度、协同决策、应急处置等方面的基础理论研究是智能调度亟待解决的难题。本项目从剖析高铁调度指挥本质和全过程着手，以运输态势评估、演化与保障为研究脉络，围绕高铁列车运行态势评估与演化理论、高铁列车晚点传播-恢复理论、高铁列车运行调整智能化理论、多工种协同高铁调度决策智能化理论四个科学问题展开研究。首先，建立了基于多源信息融合的高铁运输组织态势与调度指挥风险评估理论与方法；运用Word2vec、TF-IDF加权模型和混合神经网络模型建立了数据驱动的高铁晚点推衍模型，揭示了时间-事件混合驱动下高铁列车运行态势演化机理。第二，提出了基于BIRCH模型与梯度提升回归树的高铁晚点分类方法，构建了融合卷积、长短记忆以及全连接前馈神经网络的高铁晚点传播与预测深度学习模型；构建了高铁初始晚点恢复随机森林预测模型，揭示了高铁列车运行调整策略的晚点恢复机理；构建了基于晚点恢复效率最大化的高铁运行图冗余时间布局优化模型。第三，建立了基于知识自动化的高铁列车运行冲突检测规则及方法，构建了区间能力部分失效和全失效下基于事件-活动网络的高铁列车运行调整线性模型与有效算法；提出了基于本体概念模型的高铁列车运行调整自动化知识构建规则与方法。最后，建立了基于随机Petri网的高铁应急调度多工种协同处置流程模型及分析方法；提出了面向多工种协同的高铁列车调度、动车调度、乘务调度、施工维修调度计划两两协同优化模型与方法；研究了基于调度电话与文本信息的调度命令智能生成方法。本项目提出的高铁智能调度理论、模型与算法，在我国典型高铁线路（如京沪、武广、沪昆、成渝高铁）和车站（如上海虹桥、郑州东、西安北、成都东站）得到实例验证，部分成果为现场决策提供辅助支持。本项目建立了面向“点→线→面”态势保障的高铁调度协同决策智能化理论与方法，可为实现我国高铁智能调度、加快智能高铁建设提供理论和方法支撑。