Machine Learning-Based Estimation of Fuel Consumption and Route Optimization for Marine Vessels

Abstract

Accurate estimation of fuel consumption is vital for reducing operating costs and environmental emissions in marine
vessel operations. Conventional physics-based and empirical models often struggle to capture the complex and nonlinear interactions among engine performance, vessel characteristics, operational conditions, and environmental factors under real-world conditions. The increasing availability of onboard sensor data enables the use of data-driven machine learning approaches to address these challenges. This study proposes a machine learning–based framework for fuel consumption prediction and route optimization of marine vessels using operational and environmental sensor data. Engine and navigation measurements, including rotational speed, vessel speed, draft, and environmental parameters, are preprocessed through data cleaning and normalization to construct a reliable dataset for supervised learning. Three regression models—Artificial Neural Network, Random Forest Regression, and Gradient Boosting Regression—are developed and systematically evaluated using statistical error metrics and graphical analysis. The results show that machine learning models can accurately predict fuel consumption across diverse operating conditions, with ensemble-based methods outperforming neural networks. Gradient Boosting Regression achieves the highest prediction accuracy and robustness to noisy data. Furthermore, integrating predicted fuel consumption into route evaluation demonstrates measurable reductions in cumulative fuel usage. The proposed framework offers a practical and scalable solution for real-time fuel monitoring and fuel-efficient routing in maritime transportation.