A Rooftop PV Potential Assessment Method Based on Improved U-Net

The assessment of rooftop photovoltaic (PV) potential is critically important for grid planning. With the rapid development of deep learning and remote sensing imagery, utilizing deep learning to extract rooftop information has emerged as an ideal approach. However, remote sensing images differ from traditional natural images in that they contain more complex information and background details. Accurately segmenting rooftop portions from remote sensing images remains a significant challenge. While the U-Net model has demonstrated superior performance in the field of semantic segmentation, its encoder is relatively simple and lacks sufficient feature extraction capabilities. Therefore, this study proposed an improved U-Net model to extract rooftop information from remote sensing images. Finally, the potential rooftop PV installation capacity is calculated. This method was applied in Longhu District, Shantou City, Guangdong Province. The evaluation results indicate that the usable rooftop area in Longhu District was calculated to be 13.76 km<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, with an estimated rooftop PV installation capacity of 2876.37 MW.