lnu.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Auto Machine Learning for predicting Ship Fuel Consumption
Linnaeus University, Faculty of Technology, Kalmar Maritime Academy. (DISA)ORCID iD: 0000-0003-0372-7195
Lund university, Sweden.
2018 (English)In: Proceedings of ECOS 2018 - the 31st International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems, Guimarães, 2018Conference paper, Published paper (Refereed)
Abstract [en]

In recent years, machine learning has evolved in a fast pace as both algorithms and computing power are constantly improving. In this study, a machine learning model for predicting the fuel oil consumption from engine data has been developed for a cruise ship operating in the Baltic Sea. The cruise ship is equipped with legacy volume flow meters and newly installed mass flow meters, as well as an extensive set of logged time series data from the machinery logging system. The model is developed using state-of-the-art Auto Machine Learning tools, which optimises both the model hyper parameters and the model selection by using genetic algorithms. To further increase the model accuracy, a pipeline of different models and pre-processing algorithms is evaluated. An extensive model trained for a certain system can be used for optimisation simulation, as well as online energy efficiency prediction. As the models automatically adapt to noisy sensor data and thus function as a watermark of the machinery system, these algorithms show a potential in predicting ship energy efficiency without installation of additional mass flow meters. All tools used in this study are Open Source tools written in Python and can be applied on board. The study shows great potential for utilising large amounts of already available sensor data for improving the accuracy of the predicted ship energy consumption.

Place, publisher, year, edition, pages
Guimarães, 2018.
Keywords [en]
Ships, Auto Machine Learning, Predicting Fuel Consumption, Energy Efficiency
National Category
Energy Engineering
Research subject
Shipping, Maritime Science
Identifiers
URN: urn:nbn:se:lnu:diva-77180Scopus ID: 2-s2.0-85064184264ISBN: 9789729959646 (print)OAI: oai:DiVA.org:lnu-77180DiVA, id: diva2:1239629
Conference
ECOS 2018 - the 31st International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems, 17-21 June, 2018, Guimarães
Available from: 2018-08-17 Created: 2018-08-17 Last updated: 2021-04-26Bibliographically approved
In thesis
1. Reducing ships' fuel consumption and emissions by learning from data
Open this publication in new window or tab >>Reducing ships' fuel consumption and emissions by learning from data
2018 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

In the context of reducing both greenhouse gases and hazardous emissions, the shipping sector faces a major challenge as it is currently responsible for 11% of the transport sector’s anthropogenic greenhouse gas emissions. Even as emissions reductions are needed, the demand for the transport sector rises exponentially every year. This thesis aims to investigate the potential to use ships’ existing internal energy systems more efficiently. The thesis focusses on making existing ships in real operating conditions more efficient based logged machinery data. This dissertation presents results that can make ship more energy efficient by utilising waste heat recovery and machine learning tools. A significant part of this thesis is based on data from a cruise ship in the Baltic Sea, and an extensive analysis of the ship’s internal energy system was made from over a year’s worth of data. The analysis included an exergy analysis, which also considers the usability of each energy flow. In three studies, the feasibility of using the waste heat from the engines was investigated, and the results indicate that significant measures can be undertaken with organic Rankine cycle devices. The organic Rankine cycle was simulated with data from the ship operations and optimised for off-design conditions, both regarding system design and organic fluid selection. The analysis demonstrates that there are considerable differences between the real operation of a ship and what it was initially designed for. In addition, a large two-stroke marine diesel was integrated into a simulation with an organic Rankine cycle, resulting in an energy efficiency improvement of 5%. This thesis also presents new methods of employing machine learning to predict energy consumption. Machine learning algorithms are readily available and free to use, and by using only a small subset of data points from the engines and existing fuel flow meters, the fuel consumption could be predicted with good accuracy. These results demonstrate a potential to improve operational efficiency without installing additional fuel meters. The thesis presents results concerning how data from ships can be used to further analyse and improve their efficiency, by using both add-on technologies for waste heat recovery and machine learning applications.

Place, publisher, year, edition, pages
Växjö: Linnaeus University Press, 2018. p. 204
Series
Linnaeus University Dissertations ; 339
Keywords
shipping, energy efficiency, orc, machine learning, emissions
National Category
Energy Engineering
Research subject
Shipping, Maritime Science
Identifiers
urn:nbn:se:lnu:diva-78709 (URN)978-91-88898-22-7 (ISBN)978-91-88898-23-4 (ISBN)
Public defence
2018-12-13, B135, Landgången 4, Sjöfartshögskolan, Kalmar, 10:00 (English)
Opponent
Supervisors
Available from: 2018-11-12 Created: 2018-11-07 Last updated: 2024-02-15Bibliographically approved

Open Access in DiVA

fulltext(3702 kB)708 downloads
File information
File name FULLTEXT01.pdfFile size 3702 kBChecksum SHA-512
d4d48ae1616967125639e81da3537b5e226330f8f26941f1db2bd13f5e6f43051626b6f23bb96953a377228164fa01809c6d546d85cd267b6e67ca81762dd264
Type fulltextMimetype application/pdf

Scopus

Authority records

Ahlgren, Fredrik

Search in DiVA

By author/editor
Ahlgren, Fredrik
By organisation
Kalmar Maritime Academy
Energy Engineering

Search outside of DiVA

GoogleGoogle Scholar
Total: 710 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

isbn
urn-nbn

Altmetric score

isbn
urn-nbn
Total: 8927 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf