Deep Q-network and knowledge jointly-driven ship operational efficiency optimization in a seaport

This study addresses a ship operational efficiency optimization problem for a seaport. Given the number of planned inbound ships, the problem optimizes the inbound sequence of all ships and their speed profiles at different inbound stages. A mixed-integer nonlinear programming model is presented to minimize both the total time of ships’ port entry process (TTEP) and the total fuel consumption (TFC) of the ships. A novel deep Q-network and knowledge jointly-driven cooperative metaheuristic algorithm (DQNKD-CMA) is designed to solve the model. Experimental results based on real scenarios set in Tianjin Port demonstrate that DQNKD-CMA exhibits favorable performance in solving the problem. The proposed method improves ship inbound efficiency and reduces carbon emissions through operational measures, providing a cost-effective alternative to energy-saving equipment and alternative fuels for ship emission mitigation. This study offers a significant set of implications to shipping and port operators who face new carbon emission reduction challenges.