基于多元逻辑回归的集装箱码头集卡到达管理与调度

Trucks are the important support of in and out service at Chinese container terminals. Truck congestion has created serious environmental and social problems. Truck arrival management is to plan and control the time of deliveries and pick-ups. Its aim is to smooth truck arrivals. The available method of terminal operation is hard to solve this problem effectively. This project is based on the theories and methods of statistics and terminal scheduling. The objectives are to improve the demand forecasting of the terminal, to conquer the subjective influence of human factor, and to enhance the efficiency of both the terminal and truck companies. The research is to study the multinomial logistics regression methods based on historic data analysis for truck arrival number forecasting and truck turnaround time forecasting. More effort will be put forward on decentralized decision-making model between truck companies and terminals with interactive information. And application study will be carried out at Dayaowan International Container Terminal in Dalian and Yantian International Container Terminal in Shenzhen..This project studies the difficult problem of truck arrival management based on data analysis. It supplies a new method for truck arrival number forecasting and truck turnaround time estimation. A new tool of decentralized decision-making for simultaneous scheduling for terminal and truck companies with interactive information is developed. The research results would have certain instructional significance and be beneficial to enrich logistics system optimization at container terminals.

集卡是我国集装箱码头集疏运的重要支撑。集卡在码头的拥堵带来了严重的环境和社会问题。集卡到达管理是指通过规划与控制交、提箱时间，实现集卡的平稳到港。现有的码头运营方法难以有效解决这一难题。基于统计学和码头调度的理论与方法，以改善码头的需求预测、克服人为因素的主观性影响、提升码头与集卡公司共同的运作效率为目标，探究基于历史数据分析的多元逻辑回归针对集卡到港数量预测与在港时间预测的方法，重点研究集卡公司与码头之间信息交互下的分布式决策模型，并依托大连大窑湾国际集装箱码头和深圳盐田国际集装箱码头开展应用研究。.本项研究针对基于数据分析的集装箱码头集卡到达管理难题开展探索，为码头集卡到达数量的预测和集卡周转时间的估计提供新方法，为码头与集卡交互信息下的共同调度提供了分布式决策的新工具，研究成果具有较强的实践指导意义，有助于丰富集装箱码头物流系统优化的方法。

集卡是我国集装箱码头集疏运的重要支撑。集卡在码头的拥堵带来了严重的环境和社会问题。集卡到达管理是指通过规划与控制交、提箱时间，实现集卡的平稳到港。现有的码头运营方法难以有效解决这一难题。基于统计学和码头调度的理论与方法，以改善码头的需求预测、克服人为因素的主观性影响、提升码头与集卡公司共同的运作效率为目标，探究基于机器学习的集卡到港数量预测与在港时间预测的方法，重点研究集卡到达需求与码头服务供给之间的匹配双目标决策模型，和集卡在码头服务过程中受到内部集卡迟到、外部疫情影响的扰动管理。研究结果表明，1）在集卡日到达量的预测上，工作日、天气对集卡到达量有较强的关联，对集卡分三类的预测结果比传统的二类预测更精确；基于小波分解的预测方法能有效估计各类型集卡的在港周转时间；2）集卡交提箱的需求与内部设备配置的匹配，较大程度影响外部集卡在港周转时间；3）当外部集卡服务中受到扰动时，可以适当改变集卡服务顺序或者提升外部集卡服务的优先级，来迅速降低扰动带来的影响。.本项研究将大数据分析技术与传统运筹学方法相结合，针对基于数据分析的集装箱码头集卡到达管理难题开展探索，为码头集卡到达数量的预测和集卡周转时间的估计提供新方法，为码头资源与集卡交提箱需求的匹配提供了决策支持，对集卡服务中的扰动管理，提出了及时有效的应对策略，研究成果具有较强的实践指导意义，有助于丰富集装箱码头物流系统优化的方法。