In the realm of solar photovoltaic system image segmentation, existing deep learning networks focus almost exclusively on single image sources both in terms of sensors used and image resolution. This often prevents the wide deployment of such networks. Our research introduces a novel approach to train a network on a diverse range of image data, spanning UAV, aerial, and satellite imagery at both native and aggregated resolutions of 0.1 m, 0.2 m, 0.3 m, 0.8 m, 1.6 m, and 3.2 m. Using extensive hyperparameter tuning, we first determined the best possible parameter combinations for the network based on the DeepLabV3 ResNet101 architecture. We then trained a model using the wide range of different image sources. The final network offers several advantages. It outperforms networks trained with single image sources in multiple test applications as measured by the F1-Score (95.27%) and IoU (91.04%). The network is also able to work with a variety of target imagery due to the fact that a diverse range of image data was used to train it. The model is made freely available for further applications.

Deep learning (DL)—in particular convolutional neural networks (CNN)—methods are widely spread in object detection and recognition of remote sensing images. In the domain of DL, there is a need for large numbers of training samples. These samples are mostly generated based on manual identification. Identifying and labelling these objects is very time-consuming. The developed approach proposes a partially automated procedure for the sample creation and avoids manual labelling of rooftop photovoltaic (PV) systems. By combining address data of existing rooftop PV systems from the German Plant Register, the Georeferenced Address Data and the Official House Surroundings Germany, a partially automated generation of training samples is achieved. Using a selection of 100,000 automatically generated samples, a network using a RetinaNet-based architecture combining ResNet101, a feature pyramid network, a classification and a regression network is trained, applied on a large area and post-filtered by intersection with additional automatically identified locations of existing rooftop PV systems. Based on a proof-of-concept application, a second network is trained with the filtered selection of approximately 51,000 training samples. In two independent test applications using high-resolution aerial images of Saarland in Germany, buildings with PV systems are detected with a precision of at least 92.77 and a recall of 84.47.

Accurate and detailed spatial data on wind energy infrastructure is essential for renewable energy planning, grid integration, and system analysis. However, publicly available datasets often suffer from limited spatial accuracy, missing attributes, and inconsistent metadata. To address these challenges, this study presents a harmonized and spatially refined dataset of wind turbines in South Africa, combining OpenStreetMap (OSM) data with high-resolution satellite imagery, deep learning-based coordinate correction, and manual curation. The dataset includes 1487 turbines across 42 wind farms, representing over 3.9 GW of installed capacity as of 2025. Of this, more than 3.6 GW is currently operational. The Geo-Coordinates were validated and corrected using a RetinaNet-based object detection model applied to both Google and Bing satellite imagery. Instead of relying solely on spatial precision, the curation process emphasized attribute completeness and consistency. Through systematic verification and cross-referencing with multiple public sources, the final dataset achieves a high level of attribute completeness and internal consistency across all turbines, including turbine type, rated capacity, and commissioning year. The resulting dataset is the most accurate and comprehensive publicly available dataset on wind turbines in South Africa to date. It provides a robust foundation for spatial analysis, energy modeling, and policy assessment related to wind energy development. The dataset is publicly available.

The global energy landscape is experiencing growing challenges, with energy crises in regions such as South Africa underscoring the drive to accelerate the shift toward renewable energy solutions. This paper presents an approach for improving solar energy planning, specifically focusing on leveraging the capabilities of the ATlite software in conjunction with custom data. Using mathematical models, ATlite (which was initially developed by the Renewable Energy Group at the Frankfurt Institute for Advances Studies) is a Python software package that converts historical weather data into power generation potentials and time series for renewable energy technologies such as solar photovoltaic (PV) panels and wind turbines. The software efficiently combines atmospheric and terrain data from large regions using user-defined weights based on land use or energy yield. In this study, European Centre for Medium-Range Weather Forecasts reanalysis data (ERA5) data was modified using Kriging to enhance the resolution of each data field. This refined data was applied in ATlite, instead of utilizing the standard built-in data download and processing tools, to generate solar capacity factor maps and solar generation time series. This was utilized to identify specific PV technologies as well as optimal sites for solar power. Thereafter, a simulated power generation time series was compared with measured solar generation data, resulting in a root mean square error (RMSE) of 19.6 kW for a 250 kWp installation. This approach’s flexibility and versatility in the inclusion of custom data, led to the conclusion that it could be a suitable option for renewable energy planning and decision making in South Africa and globally, providing value to solar installers and planners.

Accurate and detailed spatial data on wind energy infrastructure is essential for renewable energy planning, grid integration, and system analysis. However, publicly available datasets often suffer from limited spatial accuracy, missing attributes, and inconsistent metadata. To address these challenges, this study presents a harmonized and spatially refined dataset of wind turbines in South Africa, combining OpenStreetMap (OSM) data with high-resolution satellite imagery, deep learning-based coordinate correction, and manual curation. The dataset includes 1,487 turbines across 42 wind farms, representing over 3.9 GW of installed capacity as of 2025. The Geo-Coordinates were validated and corrected using a RetinaNet-based object detection model applied to both Google and Bing satellite imagery. Instead of relying solely on spatial precision, the curation process emphasized attribute completeness and consistency. Through systematic verification and cross-referencing with multiple public sources, the final dataset achieves a high level of attribute completeness and internal consistency across all turbines, including turbine type, rated capacity, and commissioning year. The resulting dataset is the most accurate and comprehensive publicly available dataset on wind turbines in South Africa to date. It provides a robust foundation for spatial analysis, energy modeling, and policy assessment related to wind energy development. The dataset is publicly available for download and can be explored interactively online.