| Résumé | Accurate power prediction for a ship vessels is critical for improving operational efficiency and reducing fuel consumption, hence, shipborne emissions. This paper presents a machine learning approach for predicting the required power for ship at a given operational / environmental conditions using time series data that combines vessel telemetry, including speed and motions, with environmental data from ECMWF (European Centre for Medium-Range Weather Forecasts). Focusing on open water conditions to avoid ice interference, we apply ensemble methods, specifically Random Forest and XGBoost (Extreme Gradient Boost), and evaluate their performance using Time Series Split and Block Time Series Split techniques to handle temporal dependencies. Our models are assessed using Root Mean Squared Error, Mean Absolute Error, and R-squared. In this paper, the Canadian Coast Guard Vessel is used as a case study. The results demonstrate the effectiveness of machine learning in predicting vessel power and highlight the importance of selecting appropriate data splitting strategies to prevent data leakage. Index Terms—Machine Learning, Power Prediction, Time series split, Ice breaker vessel motion, Data leakage. |
|---|
