关键词: 集卡预约系统, 集装箱码头, 周转时间, 数据驱动, LSTM神经网络, 排队论

Abstract: Truck booking is an effective means to alleviate congestion in container terminals and their surrounding areas, and achieve balanced production of internal operating machinery in ports. In response to the objective demand of the industry for predicting the turnaround time of container trucks in ports in different time periods, it proposes a model and data driven method for predicting the turnaround time of container trucks in ports. The method transforms the turnaround time prediction into two sub problems: predicting the number of arriving vehicles and calculating the turnaround time inside the port. Among them, a data-driven double-layer LSTM (long short term memory) model is constructed for the prediction of the number of arriving vehicles; the calculation of turnover time in the port adopts a queuing model driven method. By comparing and analyzing with historical actual datasets, the experimental results show that compared to the traditional single data driven method or single model driven method, the data and model dual driven method proposed in this paper can effectively predict the turnaround time of terminal trucks, and can reduce root mean square error (RMSE) and mean absolute percentage error (MAPE) by more than 40% compared to the single data driven method or single model driven method. A more accurate prediction of truck turnaround time can provide favorable support for terminal operation planning.

Key words: truck appointment system, container terminal, turnaround time, data driven, LSTM neural network, queuing theory