| 考虑差异化客流需求的旅客列车开行综合协调优化 |
| Integrated coordination and optimization of passenger train operation scheme consideringclassified passenger demand |
| 摘要点击 20 全文点击 0 投稿时间：2025-09-19 修订日期：2026-08-28 |
| 查看/发表评论 下载PDF阅读器 |
| 中文关键词 票价制定;席位分配;客流弹性;人工神经网络;深度强化学习 |
| 英文关键词 ticket pricing; seat allocation; passenger flow elasticity; artificial neural network; deep reinforcement learning |
| 基金项目 国家社会科学基金项目 |
| 投稿方向 交通系统工程 |
| 作者 | 单位 | 邮编 | | 李冰* | 郑州大学管理学院 | 450001 | | 余倩倩 | 郑州大学管理学院 | | | 轩华 | 郑州大学管理学院 | |
|
| 中文摘要 |
| 研究一类考虑差异化客流需求的旅客列车开停决策-停站安排-动态定价-席位分配综合协调优化问题。依据票价敏感度对旅客类型进行划分,给出基于差异化旅客分类的广义出行成本计算方法,设计列车选择效用表达不同列车对旅客出行需求带来的满足程度,给出基于广义出行成本的差异化旅客分担量计算方法。在此基础上以列车总运营收益最大化为优化目标,考虑票价波动区间、票价增减趋势、席位分配判别、在车乘客数限制、列车开停和停站限制等约束,构建考虑旅客分层需求的列车开行综合优化模型,并设计嵌入Transformer的Actor-Critic网络深度强化学习过程进行求解。算法以候选列车为智能体,设置基于开停-停站-票价-席位数的状态空间,构建基于开停选择-停站选择-票价制定-席位分配的动作空间,结合优化目标构建奖励函数。引入Transformer编码器处理开停-停站-票价-席位数四类数据,利用Actor-Critic神经网络进行动作选择和优劣评价。以北京至上海高速铁路为实验对象,给出20列候选列车的初始停站方案、45个OD对的初始分类客流需求量和期望选择列车、初始票价与初始席位分配。进而利用深度强化学习算法求解得到考虑旅客分类需求的列车开行-停站-票价-席位同优化方案,并通过与初始方案比对发现综合优化方案实现了17.28%的总收益增幅。进一步对客流需求价格弹性进行敏感度分析,结果表明：当客流价格弹性系数增大时,票价上升将导致的经济型与商务型旅客的客流量及客票总收入均出现下降。经济型旅客表现出较高的价格敏感性,其客流量与客票总收入下降幅度较为显著；而商务型旅客的相应指标则变化趋缓,显示出较强的刚性需求特征；对比高差异度方案和低差异度方案,中等差异度方案在平衡服务效率与开行收益上表现最优,其以合理的停站密度、稳定的客流量与收入韧性,成为兼顾客流吸引力、收入保障及成本控制的最优方案。 |
| 英文摘要 |
| This study investigates the collaborative optimization of operation decision-stop plan-ticket pricing-seat allocation classified passenger demand. The passenger is classified by ticket price sensitivity. The calculation method of generalized travel cost based on classified passenger is provided. The train selection utility function is designed to express the satisfaction level caused by passenger selecting different train. The calculated approach of classified passenger ratio based on generalized travel cost is developed. Aiming at maximizing the total operating revenue of the passenger trains, an integrated optimization model of passenger train operation scheme considering differentiated passenger demand is provided. Some constraints, which include ticket price fluctuation ranges, ticket price developing trend, seat allocation distribution criteria, onboard passenger capacity limits, train operation criteria, and stop station criteria, are considered into the model. A deep reinforcement learning algorithm based on Actor-Critic neural network with Transformer encoder is proposed. Here the train is regarded as agent. The state space is formed by train operation decision, station stop, ticket price, and seat allocation. The action space is established by train operation decision, stop station selecting, ticket pricing, and seat allocating. The objective function is transformed into a reward function. A Transformer encoder is employed to deal with the four categories data and then an Actor-Critic double neural network is given to select and evaluate the integrated scheme including operation discrimination, station stop, ticket price, and seat allocation. The train operation data of Beijing-Shanghai high-speed railway is collected together to set up testing experiment. We give an initial scheme including stopping plans of 20 trains, classified passenger demand volumes and their preferable trains of 45 OD pairs, ticket price, and seat allocations. The proposed algorithm is used to solve the integrated optimization model of passenger train operation scheme with classified passenger demand (PTOS-CPD). Compared to initial scheme, the increasing ratio of the total train operation revenue based on improved scheme obtained by solving PTOS-CPD arrives at 17.28%. The sensitivity analysis of ticket price elasticity of passenger demand is further given. The result shows that as the ticket price elasticity coefficient increases, the increase in ticket prices will lead to a decrease in passenger flow demand and total ticket revenue for both economy-class and business -class travelers. The economy-class travelers exhibit the high price sensitivity characteristics with a significant decrease in passenger flow and total ticket revenue. The corresponding indicators of business-class travelers are no apparent trend of drastic fluctuations. It illustrates that the business-class travelers have the strong rigid travel demand characteristics. Comparing the high-differentiated scheme and the low-differentiated scheme, the medium-differentiated passenger flow elasticity coefficient scheme performs the best in balancing service efficiency and operating revenue. With a reasonable stopping station density, stable passenger flow demand, and ticket income resilience, the medium-differentiated scheme has become the optimal solution that balances passenger attraction, income security, and cost control. |
| 关闭 |
|
|
|
|
|
