Berth Scheduling Optimization Based On Dynamic Learning

With the rapid growth of container trade,competition among container terminals has become fierce.Berths are the most important strategic resources in the container terminal.The efficient use of berths is of great significance for improving the efficiency of operation.In practice,berth scheduling scheme is usually determined by the ship arrival time,the handling volumes and the ship size.However,severe meteorological and hydrological conditions may disrupt the original berth scheduling scheme.These disruptions not only lead to the late delivery of containers to shippers,but also add the difficulties for the fleet operation and scheduling.As for this,this paper studied the berth scheduling optimization of container terminals under the influence of meteorology and hydrology.This paper determined the berthing time,the berthing position and the number of cranes for ships,with the goal of minimizing the total cost of the crane usage and the ship delay.To improve the calculation accuracy of handling time under the influence of meteorology and hydrology,we proposed a handling time machine learning model considering wind,waves,rainfall and visibility.In the machine learning model,aimed at the large differences of handling efficiency with different meteorological and hydrological conditions,a machine learning method was used to calculate the handling efficiency under the influence of meteorological and hydrological conditions.Considering the meteorological and hydrological conditions can be predicted and frequently changed,the handling time should be estimated by accumulating the handling volumes per unit time.In order to solve the machine learning-based berth scheduling optimization model,an improved particle swarm optimization algorithm based on dynamic learning was designed.The particle swarm optimization algorithm helped to effectively solve the berth scheduling optimization problem,and the dynamic learning algorithm helped to realize the continuous update of the handling efficiency training results with the changes of environment and season.In addition,an algorithm combined with the neural network method and clustering method was designed,for the purpose of calculating the handling efficiency training results under the typical meteorological and hydrological conditions.Based on the practical operation parameters of container terminal,calculation examples were designed to verify the effectiveness of the model and algorithm.The results showed that the calculation deviation of handling efficiency after using machine learning method has been reduced,which proved that the handing time machine learning model is effective.Then,by comparing the proposed model with the chance constraint model and deterministic model,the effectiveness of the machine learning-based berth scheduling optimization model is verified.Finally,sensitivity analysis was carried out for the berth scheduling scheme under different meteorological and hydrological conditions.The results showed that under good weather condition,ships can departure on time by adjusting the number of cranes whether the ships arrival time is advanced or delayed.But under bad weather condition,the addition of the cranes has a limited contribution to improve handling efficiency.In this case,it is not recommended to accelerate the ships departure by adding the number of cranes.