Robust Optimization On Berth Allocation Problem Of Container Terminals

This thesis mainly focused on the robust optimization of berth allocation problem (BAP) in container terminals.Container terminals are of strategic importance in global logistics and domestic economy. While berth schedule is the basis of all operations in a port, it is very important to generate high efficiency berth planning to both the port owner and vessel companies. There are always some factors with uncertainty when generating such plans, such as the expected arrival times, which will pose great influence on the usability of berth plans。As a result, it is necessary and useful to consider the robustness in the process of berth allocation, which can assure that related operations can be executed smoothly and that the efficiency and service level will be improved.Robust optimization is a new approach in addressing data uncertainties and has been remarkable by many researchers. After a brief analysis on uncertainty optimization problems, the thesis introduces the research results of Soyster, Ben-Tal and Bertsimas, followed by the core problem, which is how to transform the uncertaity optimization model into the robust counterpart. The purpose is to transform the prior uncertain optimization problem into a tractable optimization problem.This thesis constructed a robust optimization model of berth allocation problem, which is a mixed integer programming (MIP) problem and can be hardly solved by software packages as the problem scale increases to a certain level. A new algorithm NPGA is developed to this model, incorporating the genetic algorithm into the nested partition framework, which is a newly proposed global optimization approach and iteratively use its four components to generate the optimal solution, i.e., partitioning, sampling, selecting and backtracking. In the final part, this thesis performs some computational experiments by NPGA algorithm and does some analysis on robustness of berth schedules. Basing on the results of experiments and the analysis, we draw the conclusion that the proposed robust optimization model can control the robustness of the solution by adjusting the parameters of protective level and thus it is of great practical value.