Degradation-aware Predictive Energy Management for Fuel Cell-Battery Ship Power System with Data-driven Load Forecasting
Degradation-aware Predictive Energy Management for Fuel Cell-Battery Ship Power System with Data-driven Load Forecasting
Authors: Timon Kopka, Sara Tamburello, Luca Oneto, Lindert van Biert, Henk Polinder, Andrea Coraddu Date: 2026-04-16 Paper ID: openalex:2604.14994
Summary
This paper addresses the operational cost optimization of hydrogen-based fuel cell-battery hybrid ship power systems by explicitly modeling both fuel consumption and cell degradation. The authors introduce a degradation-aware predictive energy management strategy that leverages data-driven load forecasting to optimize power dispatch. Applied to real operating data from a harbor tug, the approach significantly outperforms standard filter-based benchmarks in both fuel efficiency and component longevity across multiple prediction horizons.
Key Contributions
- Proposed a degradation-aware predictive energy management strategy for fuel cell-battery hybrid shipboard power systems.
- Integrated data-driven load forecasting to optimize power distribution over 15min and 1h horizons.
- Demonstrated up to 5.8% reduction in hydrogen consumption and 36.4% reduction in fuel cell degradation compared to filter-based benchmarks.
Open Questions & Future Work
Archivist Review
The paper presents a specific application of predictive energy management. I approved the question regarding EOL definition as it challenges a common modeling assumption in power system health management. The second question was rejected as a standard architectural extension (adding an outer control loop), which is common in control systems literature and doesn’t represent a foundational scientific bottleneck. No concepts were sufficiently novel or reusable outside this specific domain to warrant a permanent vault entry.
Approved Open Questions
- Fuel Cell EOL Definition: The EOL definition significantly impacts the cost model and maintenance scheduling; using a fixed threshold may lead to suboptimal lifetime management.
Rejected Candidates
- [open_question] Multi-horizon EMS Optimization (
multi-horizon-ems-optimization) - low_impact: The proposed future work is a standard architectural refinement rather than a fundamental unresolved bottleneck in time-series control or energy management.
Links
Metadata & Links
- url
- https://arxiv.org/abs/2604.14994
- paper_id
- 2604.14994
- paper_source
- openalex
- domain
- time-series
- tags
-
- architectures
-
- datasets
-
- concept_slugs
-
- dataset_slugs
-
- skill
- TimeSeriesSkill
- created_at
- 2026-04-19T05:44:57Z