Ενεργειακή αποδοτικότητα στις θαλάσσιες μεταφορές χρησιμοποιώντας intelligent route planning και ανάλυση με Python
View/
Keywords
AIS ; Μετεωρολογικά δεδομένα ; Random Forest ; Θαλάσσια δρομολόγηση ; Python ; ETAAbstract
The thesis presents a comprehensive data analysis system for improving energy efficiency in maritime transport through intelligent route planning. The core of the method is the integration of real AIS data with meteorological measurements (wind speed and wave height) in order to accurately predict the speed of travel (SOG) and calculate the required travel time on realistic sea routes. The dataset covers the period from 2015 to 2024 in the region with a longitude of 35° to 42° and a latitude of −77° to −70° (NE America region). The preprocessing includes data cleaning, checking for extreme weather values, mapping ship types into categories, and selecting ships that have a speed of at least 4 knots. To predict ship speed, a single Random Forest model is used for all years, with input features including geographical longitude and latitude (location), ship speed and type, wind speed, and wave height in the area, excluding ships with very low speeds and aiming to find the speed. The training and the evaluation of the model show high accuracy (MAE = approximately 1 knot) and a high accuracy rate (R2 = 85%). Detailed results are also produced per ship category, highlighting the stable behavior of the model. Next, a router was implemented to find a route between two ports using a node/edge graph, which avoids land. Based on the date of the route, local meteorological values are obtained for each leg of the route and fed into the trained Random Forest model to predict the speed, from which the travel time for each leg is calculated. The system calculates the total distance of the sea route, estimating the arrival time and average speed for the entire route, while providing color-coded maps of the predicted speed for each leg. The contribution of this work is twofold:
It demonstrates that the use of real AIS and meteorological data, in combination with a machine learning model, allows for accurate and useful speed estimates, and
(b) integrates these estimates into a practical route planner that maps realistic sea routes, providing measurable quantities (distance, arrival time, and average speed) for informed decision-making. The implementation is in the Python programming language, with a clear folder structure and storage of the trained model, facilitating reproduction and integration into operational environments.