基于动态学习的泊位调度方案优化

泊位调度方案的制定要依据在泊船舶的占用泊位情况及预计在泊时间，确定其他待泊船舶的靠泊位置与靠泊时间.然而，预计在泊时间受水文气象等多种复杂因素的影响，且影响程度是随时间动态变化的，给泊位调度方案的制定增加了难度.对此，本文采用动态学习方法对在泊时间计算函数进行更新，再基于所得函数对泊位调度方案进行优化.并设计了包含船舶在泊时间动态学习及泊位调度方案优化的并行算法，前者为后者提供更新的在泊时间计算函数，后者的实际执行结果为前者提供学习样本.通过算例对模型有效性进行了验证，结果表明：加入动态学习过程，船舶在泊时间的计算偏差得以降低；优化方案的平均在泊时间缩短2.4 h，总成本降低11.1%.

Berth Scheduling Scheme Optimization Based on Dynamic Learning

Berth scheduling scheme for the dwelling vessels in the port is thoroughly determined by the present berths occupied and the estimated handling time of those mooring vessels. However, as the estimation on berth handling time for each vessel is heavily affected by weather and tide etc. and all the influences are fluctuated dynamically by time. Thus the berth scheduling scheme seldom meets the requirements in practice. In this paper, the dynamic learning method is proposed to update the calculation function of handing time, and then the berth scheduling scheme is optimized based on the updated function. A parallel algorithm is designed where the dynamic learning for handling time and optimization for berth scheduling scheme are included. The former provides the updated calculation function for the latter, and the actual execution results of the latter provide learning samples for the former. The effectiveness of the model is verified by examples. The results represent that the calculation deviation of handling time can be reduced by adding the dynamic learning process; and in the optimization scheme, the mean value of handling time is reduced by 2.4 hours and the total cost is reduced by 11.1%.