Physics-Informed State Space Models for Day-Ahead Photovoltaic Power Forecasting
Accurate day-ahead photovoltaic (PV) power forecasting requires both strong temporal modeling and adherence to physical PV system constraints. Purely data-driven models often ignore the underlying PV conversion process and may produce predictions that are outside the operational range of a PV…
This paper introduces a physics-informed state space model for improving day-ahead photovoltaic (PV) power forecasting by integrating physical constraints with temporal modeling techniques.
What happened
Researchers developed a physics-informed state space framework (PowerMamba) that integrates a differentiable PV conversion model with a selective state space backbone. The approach uses physics-residual decomposition, where the physics module handles solar irradiance, temperature, and system constraints while the state space model learns residual corrections from numerical weather prediction data.
Why it matters
Accurate PV forecasting is critical for grid stability and energy management. The model's physical constraints reduce errors in real-world conditions, improving reliability for utility-scale solar operations.
- Physics-informed models improve forecasting accuracy and robustness
- Combining domain knowledge with machine learning enhances practical applicability
- Physical constraints are critical for real-world energy system reliability
- State space architectures offer efficient temporal modeling for energy forecasting
- Hybrid approaches outperform both pure physics and data-driven models
The bigger picture
The framework provides a more reliable forecasting solution for renewable energy systems by combining domain-specific physics with machine learning, reducing the risk of unrealistic predictions during data anomalies.
The physics-informed state space framework demonstrates significant improvements in day-ahead PV power forecasting accuracy and reliability through integration of physical constraints with advanced temporal modeling techniques.
Technical details
- Physics-residual decomposition architecture
- Differentiable PV conversion model with solar geometry and inverter limits
- State space backbone (PowerMamba) for temporal modeling
- Numerical weather prediction (NWP) input integration
- RMSE and MAE metrics for performance evaluation
FAQ
What is the main contribution of this research?
The development of a physics-informed state space framework that combines physical PV system constraints with temporal modeling to improve forecasting accuracy.
How does the model improve upon existing approaches?
By incorporating physical feasibility constraints and using a state space architecture to learn residual corrections from weather data, achieving significant error reduction compared to both physics-only and data-driven models.
Why are physical constraints important in PV forecasting?
They ensure predictions remain within operational limits of PV systems, reducing unrealistic outputs during data corruption or extreme weather conditions.
This summary and analysis were generated with AI from the original article and may contain errors. Read the original for the full story.
Read the full story
Originally published by arXiv cs.LG on Oct 14, 2026.
Comments
Sign in to join the discussion.