船舶交通量三阶段法预测模型及其应用
2021-10-08王伟余志成张泰山吉鹏
王伟 余志成 张泰山 吉鹏



摘要:为精准预测船舶交通量,基于多源数据融合,提出用于预测船舶交通量的三阶段法,分别预测交会水域、单条航道的船舶交通量,并以长江六圩水域以及苏北运河航道为例进行实证研究。结果表明,三阶段法预测模型具有较强的预测精度。预计2040年苏北运河航道货运承载能力饱和度达到17.13%,即未来苏北运河航道利用水平较为合理,但仍处于较低水平,尤其是蔺家坝—宿迁窑湾航段和宿迁窑湾—淮安航段的利用水平更低。本文预测方法和结果可为水路交通规划与调度管理提供支撑。
关键词: 水路运输; 三阶段法; 多源数据; 船舶交通量预测; 航道货运承载能力
中图分类号: U692
文献标志码: A
收稿日期: 2021-03-25
修回日期: 2021-06-29
基金项目: 江苏省社会科学基金(18GLB013);江苏省政府引导性计划(BR2020090)
作者简介:
王伟(1979—),男,湖北荆州人,教授,博士,研究方向为港航物流、绿色智慧港航、水灾害应急物流,(E-mail)13813826667@163.com
Meeting of the Waterborne Transport Division, World Transport Convention 2021 (WTC 2021)
Three-stage prediction model of ship traffic volume and its application
WANG Wei, YU Zhicheng, ZHANG Taishan, JI Peng
(College of Harbour, Coastal and Offshore Engineering, Hohai University, Nanjing 210098, China)
Abstract: In order to predict the ship traffic volume accurately, based on the fusion of multi-source data, a three-stage method is put forward to predict the ship traffic volume. The ship traffic volume in the intersection waters and a single channel is predicted, respectively. Taking Liuwei waters of the Yangtz River and the Grand Canal in the north of Jiangsu (North Jiangsu Canal, in short) for case study, the results show that the three-stage prediction model has strong prediction accuracy. The freight carrying capacity saturation of North Jiangsu Canal is predicted to reach 17.13% in 2040, that is, the utilization level of North Jiangsu Canal in the future is reasonable, but it is still at a low level, especially the utilization levels of Linjiaba-Yaowan of Suqian section and Yaowan of Suqian-Huaian section are lower. The prediction methods and results of this paper can support the waterway traffic planning and scheduling management.
Key words: waterway transport; three-stage method; multi-source data; ship traffic volume prediction; channel freight carrying capacity
0 引 言
船舶交通量預测是水路运输规划、设计和通航管理的主要依据[1],选择合适的预测方法和模型,能提升预测结果的可靠性,为未来水路运输建设和管理提供数据支撑[2]。
船舶交通量预测模型大多源自公路交通量预测模型,可划分为定性预测和定量预测,其中定量预测指使用历史数据和数学模型模拟船舶交通量发展趋势,如回归分析模型[3]、灰色模型[4-5]、支持向量机模型[6-8]、神经网络模型[9-10]等。为弥补单一预测模型的不足,学者们尝试结合几种单一预测……
