Voyage Segmentation and Propulsive Power Allocation: A Data-Driven Approach for Short Sea Shipping
Licentiatavhandling, 2025

Short-sea shipping, a sustainable alternative to land-based transport, faces strict environmental regulations and operational constraints to reduce fuel consumption, emissions, and costs. This thesis aims to minimise fuel consumption in short-sea shipping while adhering to sailing time constraints by developing a framework for optimising engine power allocation across predefined maritime routes. To address the limitations of existing power allocation methods, specifically their limited adaptability to metocean conditions, performance accuracy challenges, and long optimisation times, three approaches are examined: (1) Data-driven modelling, (2) Power allocation optimisation, and (3) Route segmentation.
The first part of the research project analyses a double-ended ferry. Here, data mining techniques were used to uncover trends in fuel consumption linked to power allocation of the ferry, revealing potential savings of up to 35% compared to actual operational data. Building on these findings, a decision support system (DSS) was developed, combining XGBoost to model fuel consumption and sailing time with Bayesian optimisation to recommend optimal engine speed and engine load. Full-scale experiments validated the DSS, achieving an average 18% reduction in the vessel’s fuel consumption
through the proposed engine power allocation strategies.
In the second half, the developed data-driven methods were combined with a novel voyage optimisation method performed in two steps. 1) Route segmentation: ship routes were segmented using the metocean score-based pruned exact linear time (MS-PELT) algorithm to identify optimal segments for engine power adjustments; 2) Engine power allocation, a scenario-based analysis grid was generated for each segment, and dynamic programming was used to determine the optimal power allocation for the voyage. The combined approach was tested on three years of data from a chemical tanker. Numerical simulations showed a 14% reduction in fuel consumption compared to measurement data, with sailing time deviations below 1%. This research demonstrates that the proposed framework significantly improves fuel efficiency in short-sea shipping while maintaining time constraints.

Bayesian optimisation

short sea shipping

power allocation optimisation

machine learning

double-ended ferry

dynamic programming

voyage segmentation.

M Room Delta
Opponent: Luis Sanchez-Heres, RISE, Sweden

Författare

Daniel Vergara

Chalmers, Mekanik och maritima vetenskaper, Marin teknik

A machine learning based Bayesian decision support system for efficient navigation of double-ended ferries

Journal of Ocean Engineering and Science,;Vol. 9(2024)p. 605-615

Artikel i vetenskaplig tidskrift

Power allocation influence on energy consumption of a double-ended ferry

Proceedings of the International Offshore and Polar Engineering Conference,;(2023)

Paper i proceeding

D. Vergara, X. Lang, M. Zhang, M. Alexandersson, and W. Mao. “Reduced environmental impact of short sea shipping through optimal engine power allocation. ” Manuscript submitted for Journal Publication, 2024.

AI-förbättrade energieffektivitetsåtgärder för optimal fartygsdrift för att minska utsläppen av växthusgaser

VINNOVA (2021-02768), 2021-10-15 -- 2024-06-30.

Drivkrafter

Hållbar utveckling

Styrkeområden

Transport

Energi

Ämneskategorier (SSIF 2025)

Maskinteknik

Utgivare

Chalmers

M Room Delta

Opponent: Luis Sanchez-Heres, RISE, Sweden

Mer information

Skapat

2025-02-26