Enhancing early-stage energy consumption predictions using dynamic operational voyage data

Abstract

The adverse human contribution to global climate change has forced the yachting industry to acknowledge the need to reduce its environmental impact due to the client's increasing pressure and potential future regulations to limit the ecological effects. Unfortunately, current real-world data presents a significant disparity between predicted and actual gathered energy consumption results. Thus, this research aims to develop an approach to accurately predict total dynamic Energy Consumption (EC) using real operation voyage data for the improved early-stage design of future yachts. A Grey-Box Modelling (GBM) solution combines: physics-based White-Box Models (WBM); and Black-Box Model (BBM) artificial neural networks to provide estimations with high accuracy and improved extrapolation capacity. The study utilizes ten months of onboard continuous monitoring data, hindcasted weather, and voyage information from a Feadship fleet yacht. Upon applying a sequential modelling methodology, predictions are compared between the three model categories, indicating propulsion and auxiliary estimates fall within 3% and 7% error of operational conditions. The study is then continued using external range datasets to evaluate the extrapolation potential. While GBM improvements are seen over the BBM, limitations were directly related to the strength between dynamic WBM input-output correlations. Ultimately, GBM's have the potential to improve both accuracy and extrapolation ability over existing WBM and BBM's; however, much is dependent on the strength of the input-output relationships.