Sun, Ruikai
2024.
Modelling and estimation of in-port ship
emissions: A scalable multi-stage data driven framework.
PhD Thesis,
Cardiff University.
Item availability restricted. |
![]() |
PDF (Thesis)
- Accepted Post-Print Version
Restricted to Repository staff only until 2 October 2026 due to copyright restrictions. Download (6MB) |
![]() |
PDF (Cardiff University Electronic Publication Form)
Restricted to Repository staff only Download (6MB) |
Abstract
As the global shipping industry advances toward carbon neutrality goals, in-port ship greenhouse gas (GHG) emissions are receiving growing attention from academia and policymakers. High-quality and scalable emission estimation methods are crucial for formulating effective emission reduction strategies and guiding the green transition of ports toward sustainable development. This thesis analyses accurate methods for estimating GHG emissions from ships in ports. First, a systematic literature review of 139 publications from 2000 to 2024 identified critical research gaps in data quality, emission metrics, modelling approaches, and strategy simulation. Building on this, case studies of the Busan, Los Angeles, and Felixstowe ports were conducted to examine the impact of data quality on emission estimation results. A Stepwise Multiple Nonlinear Regression method was developed using a global merchant ship technical parameters database to improve imputation accuracy and data completeness. Based on these findings, a scalable multi-stage data-driven emission estimation framework was constructed, using a single ship call as the minimum training unit and incorporating transfer learning techniques. This approach achieves an 87% reduction in computation time and a 90% reduction in data requirements, while maintaining estimation errors within 5%. Finally, the thesis applies the developed model to the Red Sea crisis as a case study to assess the environmental and economic impacts of different emission reduction strategies. The results indicate that reducing ship speed and increasing fleet size for a single shipping line achieves an optimal balance among delay control, cost optimisation, and emission reduction. These findings provide practical policy recommendations for shipping companies. This thesis makes significant contributions to theory, methodology, and practice. Theoretically, this thesis enhances the maritime transport field’s understanding of the current state and research gaps in in-port ship emission estimation through a systematic literature review. It introduces the "coverage rate" metric for evaluating the performance of maritime data imputation, systematically explores the impact of port congestion factors on emission estimation models and applies transfer learning techniques to port emission modelling. Methodologically, the thesis first addresses data quality issues by developing the SMNLR method. It then introduces a novel emission estimation framework based on berthing activity similarity and further enhances model performance by integrating feature selection, hyperparameter optimisation, and sliding time window techniques. Additionally, through the application of explainable AI techniques, the thesis identifies berthing waiting time and ship size as the main drivers of in-port ship emissions. The developed model has been validated for II application in port emission inventory compilation, offering robust data support for analysing port emission sources. Practically, the thesis uses the emission estimation model to quantify the increase in emissions caused by route changes during the Red Sea crisis and proposes a sensitivity analysis strategy to optimise both environmental and economic outcomes. This provides practical guidance for port management and shipping company operational decisions and offers valuable references for future research on shipping emission reduction policy design and supply chain resilience.
Item Type: | Thesis (PhD) |
---|---|
Date Type: | Completion |
Status: | Unpublished |
Schools: | Schools > Business (Including Economics) |
Date of First Compliant Deposit: | 3 October 2025 |
Last Modified: | 03 Oct 2025 11:14 |
URI: | https://orca.cardiff.ac.uk/id/eprint/181455 |
Actions (repository staff only)
![]() |
Edit Item |