Research on Carbon Intensity Prediction Method for Ships Based on Sensors and Meteorological Data

Abstract

The Carbon Intensity Index (CII) exerts a substantial impact on the operations and valuation of international shipping vessels. Accurately predicting the CII of ships could help ship operators dynamically evaluate the possible CII grate of a ship at the end of the year and choose appropriate methods to improve its CII grade to meet the IMO requirement with minimum cost. This study developed and compared five CII predicting models with multiple data sources. It integrates diverse data sources, including Automatic Identification System (AIS) data, sensor data, meteorological data, and sea state data from 2022, and extracts 21 relevant features for the vessel CII prediction. Five machine learning methods, including Artificial Neural Network (ANN), Support Vector Regression (SVR), Least Absolute Shrinkage and Selection Operator (LASSO), Extreme Gradient Boosting (XGBoost), and Random Forest (RF), are employed to construct the CII prediction model, which is then applied to a 2400 TEU container ship. Features such as the mean period of total swell, mean period of wind waves, and seawater temperature were considered for inclusion as inputs in the model. The results reveal significant correlations between cumulative carbon emissions intensity and features like cumulative distance, seawater temperature, wave period, and swell period. Among these, the strongest correlations are observed with cumulative distance and seawater temperature, having correlation coefficients of 0.45 and 0.34, respectively. Notably, the ANN model demonstrates the highest accuracy in CII prediction, with an average absolute error of 0.0336, whereas the LASSO model exhibits the highest error of 0.2817. Similarly, the ANN model provides more accurate annual CII ratings for the vessel. Consequently, the ANN model proves to be the most suitable choice for cumulative CII prediction.

Keywords

• water transport
• carbon intensity prediction of ship
• machine learning
• fuel consumption of ship
• data fusion