Abstract
Photovoltaic (PV) energy grows at an unprecedented pace, which makes it difficult to maintain up-to-date and accurate PV registries, which are critical for many applications such as PV power generation estimation. This lack of qualitative data is especially true in the case of rooftop PV installations. As a result, extensive efforts are put into the constitution of PV inventories. However, although valuable, these registries cannot be directly used for monitoring the deployment of PV or estimating the PV power generation, as these tasks usually require PV systems characteristics. To seamlessly extract these characteristics from the global inventories, we introduce PyPVRoof. PyPVRoof is a Python package to extract essential PV installation characteristics. These characteristics are tilt angle, azimuth, surface, localization, and installed capacity. PyPVRoof is designed to cover all use cases regarding data availability and user needs and is based on a benchmark of the best existing methods.
Introduction
As the global energy landscape shifts toward decentralized generation, rooftop photovoltaic (PV) installations have seen explosive growth. While mapping the location of these panels (often as geolocalized polygons) has improved, there remains a significant gap in technical metadata. Transmission System Operators (TSOs) and researchers require specific characteristics—namely tilt angle, azimuth, and installed capacity—to accurately predict power generation. Currently, this data is often missing or non-standardized. PyPVRoof addresses this by providing an open-source pipeline to estimate these values from available remote sensing data.
Methodology
PyPVRoof is designed for flexibility, offering three primary workflows based on data availability:
- Auxiliary Data Workflow: Utilizes existing partial registries to estimate tilt via a Look-Up Table (LUT) and calibrates capacity using linear regression.
- Digital Surface Model (DSM) Workflow: Leverages 3D data (LiDAR or photogrammetry) to fit a plane to the roof surface using Theil-Sen estimation, allowing for precise extraction of tilt and azimuth.
- Minimal Data Workflow: Defaults to a bounding-box heuristic for azimuth and uses user-imputed constants for tilt and efficiency when no external datasets are available.
Key Insights and Results
The package benchmarks reveal several critical findings for remote sensing in the PV domain:
- Robustness of Theil-Sen Estimation: This method significantly outperformed standard least-squares regression for 3D data because of its inherent resistance to noise and outliers in elevation models.
- Performance Baselines: LiDAR-based Theil-Sen estimation achieved a Root Mean Square Error (RMSE) of 4.38° for azimuth. Interestingly, for tilt estimation, the Look-Up Table (LUT) approach (RMSE 10.29°) currently outperforms LiDAR-based estimation (14.69°) in preliminary tests.
- Capacity Estimation: A clustered linear regression model achieved an RMSE of 0.687 kWp, refining the state-of-the-art for capacity prediction based on polygon surface area.
- Computational Efficiency: The LUT method is highly scalable, processing a single polygon in approximately 6 microseconds, making it viable for national-scale inventory processing.
Conclusion
PyPVRoof standardizes "best-in-class" methods for PV characteristic extraction into a reproducible Python environment. While the most accurate results currently depend on high-resolution LiDAR data, the package provides robust fallbacks for data-scarce regions. Future work will focus on refining these benchmarks and exploring the direct application of Theil-Sen estimation to raw LiDAR point clouds to further minimize vertical noise sensitivity.
Summary updated