Cargo Ship Fuel Burn Routing Model

ISEF Category: Engineering Technology: Statics and Dynamics

Subcategory: Naval Systems  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A cargo ship can burn far more fuel just by choosing a different route. A small change in wind, waves, or current can add real money to a voyage. That makes ship routing a live engineering problem, not just a navigation choice. You can study it with public data and your own model.

What Is It?

This project asks a simple question, how should a ship choose its path when the ocean pushes back? A route that looks shorter on a map can cost more fuel if it crosses rough seas or strong headwinds. That means the best route is not always the straightest one.

You will treat the ocean like a changing road system. NOAA gives you forecast data for wind and waves, and AIS data shows where ships actually went. AIS stands for Automatic Identification System, a public tracking signal that many ships broadcast. By comparing real ship tracks with your model’s predicted best path, you can test whether your router makes sense.

A* is a pathfinding method that searches for the cheapest route using a cost score. Reinforcement learning, or RL, is a machine learning method that learns by trying actions and getting rewards. In this project, those tools can estimate which route should use less fuel under changing marine conditions.

Why This Is a Good Topic

This is a strong science fair topic because you can measure it with real data, not guesses. You can test whether weather-aware routing predicts ship behavior better than distance-only routing, which connects directly to shipping costs and emissions. You also get room to learn real skills, such as data cleaning, GIS mapping, route scoring, and model comparison.

Research Questions

• How does adding NOAA wind data change predicted fuel burn for a trans-Pacific cargo route?
• What is the effect of wave height on the gap between shortest-distance routing and lowest-fuel routing?
• Does an A* router match real AIS ship tracks better than a distance-only baseline?
• To what extent do seasonal forecast differences change the route that minimizes estimated fuel use?
• Which route features, such as distance, headwind exposure, or wave exposure, best explain actual ship choices?
• How does a reinforcement learning router compare with A* on predicted fuel burn and route similarity?

Basic Materials

• Laptop with spreadsheet software and enough storage for geospatial data
• Internet access for NOAA forecast data and open AIS ship tracks
• Python installed with pandas, geopandas, numpy, matplotlib, and scikit-learn
• QGIS for viewing routes on a map
• Digital notebook for tracking data sources, cleaning steps, and model choices
• External hard drive or cloud storage for backups.

Advanced Materials

• Workstation with enough RAM for large geospatial joins and route simulations
• Python with geospatial, optimization, and machine learning libraries
• QGIS or ArcGIS if your lab has it
• Marine chart layers or coastline shapefiles from public sources
• NOAA wave and wind archive data
• Open AIS data for multiple trans-Pacific voyages
• Access to a university mentor or lab for model review and validation.

Software & Tools

• Python: Processes AIS data, scores routes, and runs routing models.
• QGIS: Visualizes ship tracks, forecast layers, and route comparisons on a map.
• pandas: Cleans voyage tables and joins weather data to ship positions.
• geopandas: Handles geospatial line and point data for route analysis.
• Matplotlib: Plots fuel estimates, route error, and seasonal patterns.

Experiment Steps

1. Define one route corridor and one fuel metric so your study stays focused.
2. Collect a sample of AIS voyages and match them to NOAA wind and wave data by time and location.
3. Build a baseline model that scores routes by distance only, then add weather penalties for wind and waves.
4. Design an A* cost map, then decide whether RL adds anything beyond the rule-based router.
5. Plan a comparison against the real ship track using route similarity, fuel estimate, and weather exposure.
6. Set up validation checks so you can test your model on voyages it did not train on.

Common Pitfalls

• Using raw AIS tracks without cleaning glitches, which makes a ship look like it jumped across the ocean.
• Mixing forecast times and ship times, which gives your model weather that the ship never saw.
• Comparing routes with different start and end ports, which hides whether the router really improved anything.
• Ignoring ocean currents or wave direction, which can make fuel estimates miss the main energy cost.
• Training and testing on the same voyages, which makes the router look better than it really is.

What Makes This Competitive

A stronger project goes beyond a simple map comparison. You can test several routing rules, then compare them with a clear error metric against real voyages. You can also split data by season, ship type, or ocean region to see where the model fails. That kind of careful validation shows real engineering thinking.

Project Variations

• Compare fuel-aware routing for container ships and bulk carriers on the same ocean corridor.
• Test whether adding ocean current data improves route predictions more than adding wave data alone.
• Compare A* routing with a simple shortest-path baseline and a learned RL policy on held-out voyages.

Learn More

• NOAA National Centers for Environmental Information: Search their marine and weather data sets for wind, wave, and forecast archives.
• NOAA National Data Buoy Center: Find public observations that help check forecast conditions against real ocean measurements.
• US Coast Guard AIS information pages: Learn how AIS works and what vessel tracking data can reveal.
• NASA Earthdata: Search for ocean and atmosphere data products that can support route context and validation.
• MIT OpenCourseWare, Introduction to Algorithms: Review shortest-path ideas like A* before building your router.
• Journal of Marine Science and Engineering: Search for review articles on weather routing, fuel optimization, and ship performance modeling.