Machine Learning for real-time prediction of onboard energy demand in maritime operations

Improving energy consumption efficiency on board ships has direct implications on operational costs, regulatory compliance, and environmental impact. Therefore, a data-driven approach is proposed for forecasting thermal and electrical energy demands under varying operational and environmental conditions, as well as for analysing the performance of the energy system. High-frequency operational data onboard a large cruise ship was used to develop and optimize machine learning models. The dataset was pre-processed by handling missing values and removing outliers to ensure reliability. Subsequently, a correlation analysis guided the selection of the most relevant input features. Fuel consumption predictions showed a maximum deviation of 2.7% compared to measured data, demonstrating the models practical accuracy. Interpretability was addressed through SHAP value analysis. The best-performing models were deployed within an interactive Streamlit-based dashboard capable of real-time and batch prediction of energy consumption and fuel usage. The tool provides easy-to-use interface and actionable insights for ship operators to support informed decision-making and promote energy-efficient maritime operations.