Fuel consumption prediction for passanger Ferry

This project is accepted at journal of ocean engineering in a regular paper entitled “Fuel Consumption Prediction for a Passenger Ferry using Machine Learning and In-service Data: A Comparative Study” with colaboration Pedram Agand, Allison Kennedy, Trevor Harris, Chanwoo Bae , Mo Chen, and Edward J. Park

As the importance of eco-friendly transportation increases, providing an efficient approach for marine vessel operation is essential. Methods for status monitoring with consideration to the weather condition and forecasting with the use of in-service data from ships requires accurate and complete models for predicting the energy efficiency of a ship. The models need to effectively process all the operational data in real-time. This paper presents models that can predict fuel consumption using in-service data collected from a passenger ship. Statistical and domain-knowledge methods were used to select the proper input variables for the models. These methods prevent over-fitting, missing data, and multicollinearity while providing practical applicability. Prediction models that were investigated include multiple linear regression (MLR), decision tree approach (DT), an artificial neural network (ANN), and ensemble methods. The best predictive performance was from a model developed using the XGboost technique which is a boosting ensemble approach.

In this study, we present an end-to-end approach for predicting SFE for a passenger vessel using a gray-box approach. The analysis is conducted on the vessel’s navigational operation characteristics and their physical relationships using a large set of sensor data collected over a period of approximately two years. Preprocessing techniques are applied to prepare the data for modeling, and different machine learning approaches are used and compared to find the best performing model. The goal is to provide real-time navigational aid, and the models can also be useful for vessel design and maintenance, as well as for autonomous navigation.

In addition, we find that the LR family of models is suitable for non-dynamic systems with well-defined linear parameters, and their performance depends on the availability and quality of the features. Due to the complexity of the system dynamics and the deficiency in the features, all of these linear approaches suffer from poor validation loss. Non-parametric approaches can be useful for unknown model structures, and their performance can be enhanced by using an ensemble of techniques. However, their complexity and computational time are downsides. Neural network structures, such as MLP, for complex dynamic systems can provide relatively good performance, but they require careful normalization of input data. We consider future predictions of vessel states for optimization purposes, with the goal of minimizing SFC under different weather and loading conditions. As a future direction, time-based models, such as LSTM, could be developed using instantaneous models and meta-heuristic approaches to provide navigational aid.

The final version (free trial until August 26, 2023), preprint, and Github is available in the following links: