Prediction of Ship Fuel Consumption Based on Broad Learning System

With the increasing attention of IMO to green shipping, and the increasingly strict restrictions on fuel regulatory and operating costs of shipping enterprises, no matter from the perspective of energy conservation and environmental protection or operating economy, ships should be put into actual operations in the future with lower fuel consumption and less emissions. At present, the researches and applications of maritime big data are mostly concentrated in the field of shipping schedules and cargoes. However, there are few studies focusing on the ship energy management. This paper proposes a fuel consumption prediction model based on the Broad Learning System (BLS) and the Danish RO-RO ship Ms Smyril is taken as the case ship. With the measured operation data, the fuel consumption prediction model of the ship is constructed by using data analysis and machine learning. Finally, compared with the existing fuel consumption prediction methods, it is proved that the prediction effects of this method are better. The rapidity of BLS can be used for real-time prediction of fuel consumption. When there are some mechanical failures of the ship which may cause the abnormal fuel consumption of the ship, it can help the engineers and the deck officers response quickly and address problems in time. It can also provide decision-making basis for navigation optimization.