Dataset for the FuelCast Benchmark - Benchmarking Tabular and Temporal Models for Ship Fuel Consumption

The related paper is available on Springer Nature Link: https://link.springer.com/chapter/10.1007/978-3-032-15535-1_4

The preprint is available on ArXiv: https://arxiv.org/abs/2510.08217

Dataset

The dataset was created using three main sources:

• High frequency auto logger data of important parameters from the ship, recorded by the KROHNE EcoMATE™ system.
• Environment data from Copernicus Marine Environment Monitoring Service
• Weather hintcast data from OpenMeteo

Tasks

Fuel Consumption Prediction

The main task of the FuelCast Benchmark is to predict (or better simulate) the total fuel consumption of a ship under arbitrary operational and environmental conditions.

The chosen input variables X are:

• Ship_SpeedOverGround
• Environment_SeaFloorDepth
• Weather_Temperature2m
• Weather_WindSpeed10m
• Weather_WindDirection10m
• Weather_WaveHeight
• Weather_WaveDirection
• Weather_WavePeriod
• Weather_OceanCurrentVelocity
• Weather_OceanCurrentDirection

The target variable y to predict is

• Consumer_Total_MomentaryFuel

Column Description

Index

The data was sampled using 5-minute intervals. Each interval has a unique index which acts as a timestamp equivalent.

Column	Data Type	Unit	Description
index	int64	-	The 5-minutes timestep index of the data sample starting with 0.

Ship Data

Column	Data Type	Unit	Description
Ship_SpeedOverGround	f64	m / s	The speed over ground (SOG) of the vessel.
Ship_SpeedThroughWater	f64	m / s	The speed through water (STW) of the vessel.
Ship_Heading	f64	°	The heading of the vessel, typically measured by a magnetic compass.
Ship_Bearing	f64	°	The bearing / track (direction of movement) of the vessel, calculated using GPS.
Ship_AnemometerWindDirection	f64	°	The measured wind direction (relative to the vessel speed/heading).
Ship_AnemometerWindSpeed	f64	m / s	The measured wind speed (relative to the vessel).
Ship_DraftAft	f64	m	The vessel draft aft
Ship_DraftFore	f64	m	The vessel draft fore

Consumer Data

Data from all fuel consumers like engines or boilers.

Column	Data Type	Unit	Description
Consumer_<consumer-name>_FuelType	String	-	The type of the used fuel per consumer. E.g. DM, RM 380, ...
Consumer_<consumer-name>_RotationSpeed	f64	rpm	Rotation speed of the consumer shaft. Only available for engines.
Consumer_<consumer-name>_ShaftPower	f64	W	The output power of the consumer. Only available for engines.
Consumer_<consumer-name>_MomentaryFuel	f64	kg / s	The measured fuel consumption of the consumer.
Consumer_Total_MomentaryFuel	f64	kg / s	The measured fuel consumption of all consumers together.
Consumer_Total_ShaftPower	f64	W	The output power of all engines together.

Propeller Data

Column	Data Type	Unit	Description
Propeller_<consumer-name>_ShaftPower	f64	W	The power of the propeller shaft.
Propeller_<consumer-name>_RotationSpeed	f64	rpm	The RPM of the propeller.
Propeller_<consumer-name>_ShaftTorque	f64	N m	The torque of the propeller shaft.
Propeller_Total_ShaftPower	f64	W	The power of all propeller shafts together.

Environment Data

The environment data is based on reanalysis data (Global Ocean Waves Analysis and Forecast) from Copernicus Marine Environment Monitoring Service.

Column	Data Type	Unit	Description
Environment_SeaFloorDepth	f64	m	The sea floor depth below sea level (bathymetry) from Copernicus Marine.

Weather Data

The weather data is based on reanalysis data (Historical Forecast and Marine Forecast) from satellite images and other sources. The data was taken from OpenMeteo.

Column	Data Type	Unit	Description
Weather_Temperature2m	f64	°C	OpenMeteo: The air temperature at 2 meters above ground.
Weather_RelativeHumidity2m	f64	%	OpenMeteo: The relative humidity at 2 meters above ground.
Weather_Precipitation	f64	mm	OpenMeteo: Total precipitation (rain, showers, snow) sum of the preceding hour.
Weather_WeatherCode	f64	WMO code	OpenMeteo: Weather condition as a numeric code following WMO weather interpretation codes.
Weather_SurfacePressure	f64	hPa	OpenMeteo: Atmospheric air pressure at surface level; decreases with elevation.
Weather_WindSpeed10m	f64	m/2	OpenMeteo: Wind speed at 10 meters above ground.
Weather_WindDirection10m	f64	°	OpenMeteo: Wind direction at 10 meters above ground.
Weather_WindGusts10m	f64	m/s	OpenMeteo: Maximum wind gusts at 10 meters above ground over the preceding hour.
Weather_ShortwaveRadiation	f64	W / m^2	OpenMeteo: Average shortwave solar radiation over the preceding hour.
Weather_DirectRadiation	f64	W / m^2	OpenMeteo: Average direct solar radiation on the horizontal plane over the preceding hour.
Weather_DiffuseRadiation	f64	W / m^2	OpenMeteo: Average diffuse solar radiation over the preceding hour.
Weather_DirectNormalIrradiance	f64	W / m^2	OpenMeteo: Average direct solar radiation on the normal plane over the preceding hour.
Weather_SunshineDuration	f64	s	OpenMeteo: Sum of sunshine duration within the preceding hour.
Weather_WaveHeight	f64	m	OpenMeteo: Significant mean wave height.
Weather_WaveDirection	f64	°	OpenMeteo: Mean direction of significant waves.
Weather_WavePeriod	f64	s	OpenMeteo: Period between significant waves.
Weather_WindWaveHeight	f64	m	OpenMeteo: Significant wind wave height.
Weather_WindWaveDirection	f64	°	OpenMeteo: Mean direction of wind waves.
Weather_WindWavePeriod	f64	s	OpenMeteo: Period between wind waves.
Weather_WindWavePeakPeriod	f64	s	OpenMeteo: Peak period between wind waves.
Weather_SwellWaveHeight	f64	m	OpenMeteo: Significant swell wave height.
Weather_SwellWaveDirection	f64	°	OpenMeteo: Mean direction of swell waves.
Weather_SwellWavePeriod	f64	s	OpenMeteo: Period between swell waves.
Weather_SwellWavePeakPeriod	f64	s	OpenMeteo: Peak period between swell waves.
Weather_OceanCurrentVelocity	f64	m / s	OpenMeteo: Ocean current velocity considering all components.
Weather_OceanCurrentDirection	f64	°	OpenMeteo: Direction of the ocean current flow.

Citation

@InProceedings{10.1007/978-3-032-15535-1_4,
    author="Viga, Justus
    and Mueck, Penelope
    and L{\"o}ser, Alexander
    and Weis, Torben",
    editor="Lemaire, Vincent
    and Ifrim, Georgiana
    and Bagnall, Anthony
    and Malinowski, Simon
    and Sch{\"a}fer, Patrick
    and Tavenard, Romain",
    title="FuelCast: Benchmarking Tabular and Temporal Models for Ship Fuel Consumption",
    booktitle="Advanced Analytics and Learning on Temporal Data",
    year="2026",
    publisher="Springer Nature Switzerland",
    address="Cham",
    pages="54--69",
    abstract="In the shipping industry, fuel consumption and emissions are critical factors due to their significant impact on economic efficiency and environmental sustainability. Accurate prediction of ship fuel consumption is essential for further optimization of maritime operations. However, heterogeneous methodologies and limited high-quality datasets hinder direct comparison of modeling approaches. This paper makes three key contributions: (1) we introduce and release a new dataset (https://huggingface.co/datasets/krohnedigital/FuelCast) comprising operational and environmental data from three ships; (2) we define a standardized benchmark covering tabular regression and time-series regression (3) we investigate the application of in-context learning for ship consumption modeling using the TabPFN foundation model - a first in this domain to our knowledge. Our results demonstrate strong performance across all evaluated models, supporting the feasibility of onboard, data-driven fuel prediction. Models incorporating environmental conditions consistently outperform simple polynomial baselines relying solely on vessel speed. TabPFN slightly outperforms other techniques, highlighting the potential of foundation models with in-context learning capabilities for tabular prediction. Furthermore, including temporal context improves accuracy.",
    isbn="978-3-032-15535-1"
}