A novel, data-driven heuristic framework for vessel weather routing. (1st February 2020)

• Record Type: Journal Article
• Authors: Gkerekos, Christos; Lazakis, Iraklis
• Is Part Of: Ocean engineering. Volume 197(2020)
• Publication Date: 2020-02-01
• Subjects: Weather routing -- Fuel oil consumption prediction -- Ship energy efficiency -- Machine learning; Ocean engineering -- Periodicals; Ocean engineering; Periodicals; 620.4162
• DOI: 10.1016/j.oceaneng.2019.106887
• Languages: English
• ISSNs: 0029-8018

Abstract

Abstract: Fuel Oil Consumption (FOC) constitutes approximately two-thirds of a vessel's voyage costs and profoundly correlates with the adversity of the weather conditions along its route. Furthermore, increased FOC also leads to increased emissions. As shipping is turning page towards a greener, more sustainable future, it is crucial to leverage key insights from past routes in order to identify approaches that minimise both the financial cost of operations and their Green House Gas (GHG) footprint. This study presents a novel framework for vessel weather routing based on historical ship performance and current weather conditions at a discretised grid of points in conjunction with a data-driven model that can predict main engine FOC. Subsequently, a modified version of Dijkstra's algorithm that has been fitted with heuristics is applied recursively until an optimal route is obtained. The efficacy of the proposed framework is demonstrated through a case study concerning the optimal route selection for a 160, 000 tonne DWT crude oil tanker sailing between the Gulf of Guinea and the Marseille anchorage. In this case study, an R 2 of 89.4% was obtained while predicting the vessel's FOC and five optimal routes were identified and ranked for two sailing speeds corresponding to different operating profiles, i.e. ballast and fully loaded. Highlights: The initial process of acquiring, pre-processing and analysing a dataset containing raw sailing measurements from a ship is elaborated. The process to derive a data-driven Deep Neural Network model that can adequately represent the fuel oil consumption (FOC) of a vessel under varying conditions is included. The FOC model of is utilised as part of a route optimisation process that combines the data-driven aspect of the FOC with a modification of Dijkstra's algorithm to allow for time-dependent route optimisation. KPIs pertinent to ship efficiency and performance are also suggested to enhance the route optimisation process and help identify when maintenance is required. The above are verified and validated through a case study based on actual ship data.