Research On Ship Carbon Emission Prediction Based On AIS Data Mining

The boom in the ocean carriage makes China rank top in the ocean freight volume and ports throughput in the recent years.The tangible economic benefits increased significantly with the sacrifice of air quality at the same time.The maritime carbon emissions mainly derive from the vessels,and the research on the vessel information could help us find their inner relationship.Currently,Automatic Identification System(AIS)is one of the most leading sources of the vessel information,and has been widely deployed on the varied kind of the vessels,In particular,international vessels which weigh over 300 tons,domestic cargo vessels which weigh over 500 tons and the cruise vessels are forced to equip AIS.The considerable AIS data are generated every day and their potential value is worth fully exploiting.In this thesis,we carry out the research on vessel carbon emission prediction.We focus on exploring the relationship between AIS data and the vessels carbon emissions,discovering the law and the trend of the vessel carbon emission and further designing and implementing a software system to make the prediction visualized in a map.The contributions of this thesis are summarized as follows.First,the vessels' AIS data are preprocessed to pave way to the function of the vessel carbon emission prediction.The conventional AIS data,for various reasons,may result in the missing or deviation of the vessels" trajectory data,or the outlier data,etc.Therefore,for boosting the data mining quality,this thesis cleans and converts the data first,to ensure the consistency,accuracy,integrity and accountability of the data.Then,the vessel carbon emission prediction algorithms are proposed,which is based on the AIS data mining.This thesis adopts the Back Propagation(BP)neural network,the Genetic Algorithm-Back Propagation(GA-BP)neural network and the Genetic Algorithm-Least Squares Support Vector Machine(GA-LSSVM)model to respectively predict the vessels' carbon emissions.We further contrasts the predictable performance,or the above proposed vessel carbon emission preaiction algorithms.Finally,we design and impiement a AIS data mining-based vessel carbon emission prediction system.It is comprised of the user interface and management module,AIS data preprocessing module,the vessel carbon emission prediction algorithm selection and execution model,and the thermodynamie diagram display module.The system can make the association between the carbon emissions and the other modules,to form the prediction results into a visual map,which could provide a theoretical guidence and decision-making basis for the department concerned.