EventsThe 2nd International Online Conference on Marine Science and Engineering
Published
This submission belongs to the session S1. Ocean Engineering of the event The 2nd International Online Conference on Marine Science and Engineering
Published date
18 Nov, 2026
Academic Editor
author-avatarWeicheng Cui
Citation
Hamed Majidiyan, Gholam Reza Emad, T M Rabiul Islam, james Curr, An Enhanced Physics-Guided Machine Learning Model for Estimating Ship Fuel Consumption from AIS Data, in Proceedings of The 2nd International Online Conference on Marine Science and Engineering, 23 November–25 November 2026, MDPI: Basel, Switzerland
Share
Email
Facebook
Twitter
LinkedIn

An Enhanced Physics-Guided Machine Learning Model for Estimating Ship Fuel Consumption from AIS Data

image
image
image
james Curr 3
1. Australian Maritime College, University of Tasmania, Launceston, Australia, 7248.
2. Centre for Seafaring and Maritime Operations, Australian Maritime College, University of Tasmania, Newnham Campus, Launceston, Australia, 7248.
3. Centre for Maritime Engineering and Hydrodynamics, Australian Maritime College, University of Tasmania, Newnham Campus, Launceston, Australia, 7248.
Abstract

Shipping is a major pillar of global transportation and remains a significant contributor to greenhouse-gas (GHG) emissions. In response, international regulations and industry decarbonisation targets have increased the need for reliable ship-level fuel-consumption and estimation methods to calculate the downstream GHG emissions. Automatic Identification System (AIS) data are widely used in bottom-up emission inventories because they provide large-scale information on vessel movement, speed and operational patterns. However, many existing AIS-based approaches rely on simplified load-based or resistance-based formulations, which may not sufficiently capture the effects of vessel characteristics, operating conditions and environmental loads. To address this, the current study proposes an enhanced physics-guided machine learning framework for estimating ship fuel consumption from AIS-derived operational data. The proposed model incorporates hydrodynamic physics through resistance-based features, including calm-water resistance and added resistance in waves. A broad parametric dataset was developed for Supramax bulk carriers operating under representative Atlantic sea-state conditions. The hydrodynamic response matrix was enhanced using Maxsurf-based simulations, allowing the model to better account for ship geometry, speed, draft, wave height, wave period and wave heading effects. These physics-informed outputs were then used to train a neural-network model for fuel-consumption estimation. The trained model was further evaluated using real-world voyage data. The results indicate that the proposed Maxsurf-enhanced physics-guided machine learning model improves fuel-consumption estimation compared with other approaches. By combining AIS data, hydrodynamic modelling and machine learning, the framework provides a practical pathway for more accurate voyage-based fuel-consumption estimation and downstream greenhouse-gas emission assessment. The proposed approach can support ship operators, researchers and regulators in improving monitoring, reporting and verification of maritime emissions under real operating conditions.

Keywords
AIS
ship fuel consumption
greenhouse-gas emissions
physics-guided machine learning
Maxsurf
added resistance
bulk carrier
emission inventory
decarbonisation
VIBRATION DATA-QUALITY ASSESSMENT FOR REFRIGERATED SHIPPING CONTAINERS IN MARITIME COLD-CHAIN OPERATIONS
Physics-Informed Neural Networks (PINNS) for Real-Time Roll Prediction