Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Department of Logistics and Maritime Studies | en_US |
| dc.contributor.advisor | Wang, Shuaian Hans (LMS) | en_US |
| dc.creator | Chu, Zhong | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/13844 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | en_US |
| dc.rights | All rights reserved | en_US |
| dc.title | Machine learning in port operations : evaluation, prediction, and optimization | en_US |
| dcterms.abstract | Port operations are critical to global trade but face challenges like surging traffic and dynamic coordination needs. Addressing these complexities requires innovative analytical approaches that can effectively measure performance, forecast vessel activities, and optimize resource allocation. Inspired by the machine learning (ML) framework of evaluation, prediction, and optimization, this thesis explores the application of ML and operations research (OR) in port operations, with a focus on vessel arrival and departure. In the evaluation phase, two novel data fusion approaches are introduced to quantify the operational status of vessel movements, providing a more comprehensive assessment of arrival and departure dynamics. The first integrating the vessel estimated time of arrival (ETA), actual time of arrival (ATA), and the corresponding data from the Automatic Identification System (AIS) to quantify vessel arrival time (VAT) delays. The analysis reveals that as the vessels approach their destination port, their reported ETA becomes increasingly accurate in both spatial and temporal dimensions. The second study integrates the vessel's estimated departure time (EDT), actual departure time (ADT), and berth entry/exit timestamps to quantify vessel turnaround time (VTT) and service time (VST). A quantitative analysis is conducted to evaluate the impact of COVID-19 on port operations, with Hong Kong Port as a case study. The findings indicate that COVID-19 and its restrictions worsened vessel arrival delays and extended turnaround time, reducing port efficiency. | en_US |
| dcterms.abstract | The prediction phase focuses on estimating VAT, VTT, and VST. Using the established evaluation framework, relevant datasets are constructed to enable time prediction via tree-based models. This thesis is the first to integrate vessel-reported ETA and AIS data for VAT prediction of oceangoing vessels. Compared to vessel-reported ETA, the proposed approach lowers the mean absolute error (MAE) from 6.84 to 3.11 hours, at a 54.53% reduction. For inland waterway shipping, vessel traffic flow data and the A-Star algorithm are integrated to account for river transport characteristics and estimate the remaining sailing distance for VAT prediction. Results indicate a significant improvement, reducing MAE from 17.06 to 3.49 hours: a 79.54% reduction. For VTT prediction, the proposed model enhances accuracy, lowering MAE from 5.12 to 3.94 hours compared to vessel-reported values. Likewise, for VST prediction, MAE decreases from 4.54 to 3.19 hours. | en_US |
| dcterms.abstract | The optimization phase examines the impact of integrating VAT predictions into berth allocation planning (BAP). Leveraging the predicted VAT, a two-stage prediction-then-optimization framework is proposed. In the first stage, a VAT prediction model improves the accuracy of VAT estimates. In the second stage, the predicted VAT is incorporated into the BAP model to optimize berth scheduling. The effectiveness of VAT-based scheduling is evaluated by comparing a BAP model using predicted VAT with another based on vessel-reported ETA in both discrete and continuous berth settings. In a discrete berth scenario with 12 vessel arrivals, VAT-based scheduling reduces additional BAP costs by 64% and vessel waiting time by 73% compared to ETA-based scheduling. In a continuous berth setting, VAT-based scheduling reduces additional BAP costs by 43% and vessel waiting time by 35%. These findings highlight the effectiveness of VAT-based scheduling in improving berth allocation, reducing vessel waiting time, and optimizing resource utilization. | en_US |
| dcterms.abstract | By systematically incorporating data-driven insights into decision-making, this study highlights the significant potential of AI-powered port management in optimizing daily port operations through vessel arrival/departure prediction models and dynamic berth scheduling optimization, advances maritime digitalization via AI-enhanced terminal operating systems with real-time nautical data integration, and accelerates decarbonization efforts through emission-aware vessel sequencing algorithms and predictive shore power allocation. | en_US |
| dcterms.extent | xvi, 216 pages : color illustrations | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2025 | en_US |
| dcterms.educationalLevel | Ph.D. | en_US |
| dcterms.educationalLevel | All Doctorate | en_US |
| dcterms.accessRights | open access | en_US |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/13844