基于状态空间的连续断面船舶交通流量预测研究

船舶流量预测是船舶交通流研究的重要内容,建立科学合理的船舶流量预测模型有助于航道的设计、规划和管理。将传统的单断面船舶交通流预测方法向多断面进行改进和推广,提出基于状态空间和卡尔曼滤波的多断面交通流预测模型。利用船舶交通流多断面流量数据的时间序列进行多维线性回归,并转化为状态空间模型形式;在此基础上由卡尔曼滤波算法对交通流量进行递推预测,得到多断面交通流的预测值。作为实证研究,分别对武汉长江大桥、武汉长江二桥2个断面,以及长江重庆段朝天门、万州、巫山3个断面进行实际数据分析来验算模型的有效性,并与单断面多维线性回归预测方法进行对比。结果表明,使用状态空间模型得到的武汉长江大桥、二桥预测结果的平均相对误差分别减少4.59%,0.97%;而重庆段3个连续观测点采用状态空间法预测比使用时间序列预测平均绝对误差和平均相对误差均有不同程度的降低,其中平均相对误差分别降低1.08%,4.28%, 3 .54%。因此,在不同时间维度上,该模型有助于提高多断面交通流预测精度。

A Study of Vessel Traffic Flow Forecast Based on State Space Analysis of Continuous Cross Sections

The vessel traffic flow is the foundation of decision makings related to the design ,planning and manage‐ment of channels .This paper investigates the traditional vessel traffic flow forecast model of single cross section in water‐way ,and studies continuous cross section model based upon state space and Kalman filtering .The space state model is de‐veloped based on a multi‐dimensional linear regression analysis of the time series formed from the traffic flow at the multi‐ple cross sections .Kalman filtering is used to recursively forecast the vessel traffic flow at the multi cross sections .In the study ,the vessel traffic flows at the Wuhan Yangtze Bridge and Wuhan Yangtze No .2 Bridge in the City of Wuhan ,Chi‐na were analyzed to test the proposed model .The result shows the error decreases 4 .59% and 0 .97% respectively ,com‐paring to the results from the traditional forecast model using single cross section .In addition ,the real vessel traffic flow collected at Chaotianmen ,Wanzhou and Wushan in Chongqing ,China were also used to compare the proposed model with the single cross section model ,which results a 1 .08% ,4 .28% ,and 3 .54% decreases respectively .The test results indi‐cate that the proposed model can provide more accurate vessel traffic forecast .