LOSCO A Ship Lock Scheduling Model

Abstract

A ship lock scheduling model called LOSCO was developed in order to decrease passage times at locks with two parallel chambers. Three locks were chosen to model, as they are some of the busiest locks of the Netherlands. The Krammersluizen, the Sluizen Hansweert and the Kreekraksluizen. Passage times are not collected at the locks, therefore the model was compared to SIVAK. SIVAK is the standard model for research on locks at Rijkswaterstaat. To schedule vessels in locks three types of choices can be made. Every vessel needs to be assigned to a chamber, the initiation time of locking should be decided and the order of sailing in a chamber should bedefined. By optimising these choices, passage times can be decreased. Optimising is however not straightforward, as the problem is an instance of the job shop scheduling problem. No exact algorithm has been found to solve these problems in a practical amount of time. Therefore the challange is to find a balance in solution quality and the speed of the model when creating a lock scheduling model. In this thesis, four ideas to improve the lock scheduling model by Verstichel were researched. The first idea is to change the resolution of the timesteps. The gain in performance was however not found to weight up against the loss in solution quality for resolution to be useful. The second idea is to drop the first come, first serve constraint that Verstichel created. This idea was also not found to be effective. The third idea is to divide the scheduling problem up into chunks. This is called cut separation. Chunks of around 25 vessels were found to be effective. The fourth and last idea is that of a maximum waiting time. The maximum waiting time makes the performance of the model better and also makes the scheduling model fairer. Data from the year 2016 was used to schedule all lockings. The LOSCO model is effective for reducing the average assage time per vessel with about 3.9 ± 0.12 SE minutes for the Krammersluizen and 2.0 ± 0.14 SE minutes for the Sluizen Hansweert compared to the SIVAK model. At the Kreekraksluizen, the model could not find a solution, as the Kreekraksluizen are a lot busier than the other locks that were tested. At the Kreekraksluizen on average every 7.8 minutes a vessel arrives, whereas at the Krammersluizen and the Sluizen Hansweert respectively every 14.0 and 13.0 minutes a vessel arrives throughout the year. The LOSCO model is only better than SIVAK if the lock is relatively quiet. At the busiest time of the day, typically in the afternoon, SIVAK performs better. The models perform equal at inter arrival times of around 8
to 10 minutes. Optimisation on economical value of the vessels was found to be less effective than optimisation on time. Optimisation on time was also found to be fairer. The LOSCO model is a step ahead towards a practical lock scheduling model. In order to achieve a fully practical model, some simplifications need to be expanded. It is recommended to first improve the model before it is applied in practice, as the model is able to outperform SIVAK in some cases, but not in the busiest cases. After this some extra features can be implemented, such as the model dealing with vessel delays and locks with 3 chambers.