File(s) under permanent embargo
An application of machine learning to shipping emission inventory
journal contribution
posted on 2023-05-19, 23:14 authored by Fletcher, T, Vikrambhai GaraniyaVikrambhai Garaniya, Shuhong ChaiShuhong Chai, Abbassi, R, Yu, H, Van, TC, Brown, RJ, Khan, FThe objective of this study is to develop a shipping emission inventory model incorporating Machine Learning (ML) tools to estimate gaseous emissions. The tools enhance the emission inventories which currently rely on emission factors. The current inventories apply varied methodologies to estimate emissions with mixed accuracy. Comprehensive Bottom-up approach have the potential to provide very accurate results but require quality input. ML models have proven to be an accurate method of predicting responses for a set of data, with emission inventories an area unexplored with ML algorithms. Five ML models were applied to the emission data with the best-fit model judged based on comparing the real mean square errors and the R-values of each model. The primary gases studied are from a vessel measurement campaign in three modes of operation; berthing, manoeuvring, and cruising. The manoeuvring phase was identified as key for model selection for which two models performed best.
History
Publication title
Royal Institution of Naval Architects. Transactions. Part A. International Journal of Maritime EngineeringVolume
160Issue
A4Pagination
A381-A396ISSN
1479-8751Department/School
Australian Maritime CollegePublisher
Royal Institution of Naval ArchitectsPlace of publication
United KingdomRights statement
Copyright 2018 The Royal Institution of Naval ArchitectsRepository Status
- Restricted