A Machine Learning-Based Approach for Designing SEEMP on Ships: Case Study of CO₂ Emissions at a Container Port

Elwas Cahya Wahyu Pribadi (Department of Mechanical Engineering, Lambung Mangkurat University, Banjarmasin 70123, Indonesia, and is also a Doctoral Student at the Department of Marine Engineering, Institut Teknologi Sepuluh Nopember, Surabaya 60111, Indonesia.)
Abdul Ghofur (Departement of Mechanical Engineering, Universitas Lambung Mangkurat, Banjarmasin, 70123, Indonesia.)
Rachmat Subagyo (Departement of Mechanical Engineering, Universitas Lambung Mangkurat, Banjarmasin, 70123, Indonesia.)
Ma’ruf (Departement of Mechanical Engineering, Universitas Lambung Mangkurat, Banjarmasin, 70123, Indonesia.)
Akhmad Syarief (Departement of Mechanical Engineering, Universitas Lambung Mangkurat, Banjarmasin, 70123, Indonesia.)
Aldinor Setiawan (Departement of Mechanical Engineering, Universitas Lambung Mangkurat, Banjarmasin, 70123, Indonesia.)

International Journal of Marine Engineering Innovation and Research
Vol. 10 No. 4 (2025)

The management of ship energy significantly affects both cost efficiency such as earnings before interest and environmental sustainability, particularly in reducing CO₂ emissions produced by ship operations. Despite the importance of this issue, research on ship energy consumption within container terminals remains limited. This study aims to estimate CO₂ emissions generated by ship activities in container ports, focusing on emission variables related to the Ship Energy Efficiency Management Plan (SEEMP). The calculation considers active ship movements in the port, including approach, maneuvering, and berthing processes. Energy consumption and CO₂ emissions were analyzed using random forest regression (RF) with default settings, and the model’s accuracy was validated through k-fold cross-validation. The results identified five major factors influencing CO₂ emissions: (1) main engine power, (2) auxiliary engine power, (3) waiting time in the port, (4) maneuvering time, and (5) berthing time. Among these, maneuvering, waiting, and berthing showed the highest significance, confirmed by attribute selection and validation results. The random forest model demonstrated a prediction accuracy of 98.89%, confirming its reliability. Moreover, operational fuel efficiency analysis indicated that combining voyage optimization, skilled operators, and cold ironing facilities could reduce CO₂ emissions by up to 20%. These findings provide valuable insights and serve as a foundation for developing a more effective Ship Energy Efficiency Management Plan to enhance environmental performance in maritime operations.