Port carbon emission estimation: Principles, practices, and machine learning applications

Abstract

Ports are central to global trade and transportation, playing a significant role in worldwide carbon emissions. Estimating carbon emissions from ports is crucial for identifying emission sources and devising strategies for their reduction. The procedure of port carbon emission estimation follows the logic of “estimation methods — data acquisition and preprocessing — application realization” and faces many challenges currently. First, this review explores the complex terrain of port carbon emission estimation, addressing both ship-side and shore-side emissions, which examines the principles and applications of these methods. Second, evaluates the role of machine learning (ML) technologies in enhancing data accuracy due to the low quality of raw acquisition data. Specifically, Generative Adversarial Networks (GAN) proves useful in repairing ship-side and port-side raw production data. Finally, to support port authorities and government decision-makers in carbon emission estimation realization, the development of effective and practical software applications is essential, which follows a logical sequence: “conceptual design — prototype design — improved design”. This review focuses on data-driven approaches for assessing port carbon emissions while acknowledging potential limitations, such as those associated with sensor-based estimation techniques. Future research should compensate for this shortcoming by refining sensor calibration techniques and integrating complementary data sources to enhance the accuracy and reliability of port carbon emissions estimates.