Time Series Analysis for Digital Twins in Green Shipping

Abstract

A promising way for the waterborne industry towards decarbonization and emissions reduction is through digitalization and in particular via Digital Twins (DTs) technology. In this context, the DT provides insights for optimal decision-making, predicting potential future events, or even detecting irregularities in the behavior of the ship to reduce its carbon emissions and energy consumption. To achieve this, we propose an architecture for automated data capture, processing, and analysis. The analysis component of this infrastructure leverages machine learning (ML) algorithms for time series data, such as anomaly detection and forecasting. Importantly, to understand how these algorithms make a certain prediction we also provide a detailed look at current approaches used to interpret these models. Finally, we demonstrate a practical use case, where time series analysis can prove especially useful when applied to real-world vessel data.