Data-driven modelling of ship propulsion and the effect of data pre-processing on the prediction of ship fuel consumption and speed loss

Data-driven models for ship propulsion are presented while the effect of data pre-processing techniques is extensively examined. In this study, a large, automatically collected with high sampling frequency data set is exploited for training models that estimate the required shaft power or main engine fuel consumption of a container ship sailing under arbitrary conditions. Emphasis is given to the statistical evaluation and preprocessing of the data and two algorithms are presented for this scope. Additionally, state-of-the-art techniques for training and optimizing Feed-Forward Neural Networks (FNNs) are applied. The results indicate that with a delicate filtering and preparation stage it is possible to significantly increase the model's accuracy. Therefore, increase the prediction ability and awareness regarding the ship's hull and propeller actual condition. Furthermore, such models could be employed in studies targeting at the improvement of ship's operational energy efficiency.

Automated summary

This study presents data-driven models for ship propulsion, focusing on the statistical evaluation and preprocessing of data. By utilizing high frequency data sets and advanced algorithms, the accuracy of the models can be greatly improved, enhancing prediction ability regarding the ship's actual condition.