FuelCast: Benchmarking Tabular and Temporal Models for Ship Fuel Consumption

2025-10-09

AI Summary

Accurate prediction of ship fuel consumption is critical for improving maritime energy efficiency and reducing emissions; however, fair model comparison remains hindered by methodological heterogeneity and scarcity of high-quality, standardized datasets. To address this, we introduce the first publicly available multi-vessel operational–environmental fused time-series dataset and establish a unified benchmark. We pioneer the application of in-context learning (ICL) to ship energy consumption forecasting, building upon the lightweight TabPFN foundation model while integrating vessel speed, environmental variables, and temporal features. Comprehensive experiments demonstrate that incorporating environmental and temporal context significantly enhances prediction accuracy. TabPFN achieves competitive performance—slightly surpassing conventional regression models—while maintaining low computational overhead and real-time inference capability, confirming its suitability for onboard deployment. This work provides a reproducible benchmark and a novel paradigm for data-driven intelligent energy efficiency optimization in maritime operations.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Planning, Routing, and Scheduling: Planning with Language Models
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

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.

Problem

• Predicting ship fuel consumption accurately for maritime optimization
• Addressing heterogeneous methodologies hindering model comparison
• Investigating in-context learning for tabular and temporal prediction

Innovation

• Released new ship operational dataset for benchmarking
• Applied TabPFN foundation model with in-context learning
• Incorporated temporal context to improve prediction accuracy