Integrated multi-method approach for ship power prediction

Accurate power prediction is crucial for optimizing fuel consumption models in maritime vessels, particularly as the industry shifts towards more sustainable energy solutions. Traditional reliance on fossil fuels often leads to inefficiencies in large-scale ship engines, prompting a growing interest in Liquefied Natural Gas (LNG) as a cleaner alternative for propulsion systems. This study presents a comparative analysis of three approaches to power prediction: physics-based, data-driven, and a hybrid multi-method framework. The proposed integrated approach combines Computational Fluid Dynamics (CFD), grounded in first-principle equations, with machine learning techniques to better capture the complexity of real-world maritime operations. A deep learning gating mechanism is employed to dynamically select the most accurate output from the ensemble of models, enhancing both reliability and precision in power estimation. The hybrid framework offers a robust and adaptable solution that contributes to improved fuel efficiency and the reduction of greenhouse gas emissions in the maritime sector.