AI and Open-data Driven Scalable Solar Power Profiling

Problem
This preprint addresses the significant gap in the availability of detailed, up-to-date spatial data on rooftop solar photovoltaic (PV) systems. Despite the rapid expansion of solar PV deployment, existing datasets are often limited, proprietary, and require extensive manual labeling. The authors propose a scalable framework that utilizes open data to detect solar panels and generate comprehensive city-level solar power profiles, thereby democratizing access to solar energy data.

Method
The core technical contribution of this work is a framework that employs foundation vision AI models to detect solar panel geometries from open-source satellite imagery. The architecture leverages state-of-the-art object detection techniques, although specific model architectures are not disclosed. The method circumvents the need for manual data labeling and case-specific model training, enhancing robustness across diverse imagery sources. Detected solar panels are transformed into georeferenced polygons, creating spatially explicit inventories that can be incrementally updated. The framework integrates open weather data to convert panel footprints into regional solar power profiles. The authors also provide an API that allows users to input specific building locations, retrieve aerial imagery, detect rooftop solar panels, and obtain georeferenced polygons, facilitating further analysis.

Results
The authors demonstrate the effectiveness of their framework through various case studies, although specific quantitative results (e.g., precision, recall, F1 scores) against named baselines are not provided in the abstract. The framework’s ability to generate accurate solar power profiles is implied to be significant, as it enables advanced analyses such as distributed solar production integration and local power flow optimization. The authors claim that their approach significantly reduces reliance on proprietary data sources and manual processes, although exact metrics quantifying this reduction are not detailed.

Limitations
The authors acknowledge that their framework’s performance may vary based on the quality and resolution of the satellite imagery used. Additionally, the reliance on open weather data may introduce variability in solar power profiling accuracy, depending on the granularity and temporal resolution of the weather data. The paper does not address potential biases in the satellite imagery or the limitations of the foundation models used for detection, which could affect the generalizability of the results across different geographic regions.

Why it matters
This work has significant implications for solar energy research and urban planning. By providing a transparent and scalable method for solar panel detection and profiling, it enables researchers and developers to conduct advanced analyses that can inform energy policy, infrastructure planning, and local energy management strategies. The open-access nature of the data and API fosters collaboration and innovation in the field, potentially accelerating the transition to renewable energy sources and enhancing the integration of distributed solar power into existing energy systems.

Authors: Shiliang Zhang, Sabita Maharjan, Damla Turgut
Source: arXiv:2605.02738
URL: https://arxiv.org/abs/2605.02738v1