Estimation of photovoltaic power generation in traditional protected villages in mountainous areas based on satellite image semantic segmentation and 3D terrain reconstruction

With the growing demand for renewable energy, rooftop PV systems have gained widespread attention and adoption. However, existing methods for assessing PV potential are designed for urban areas and cannot address traditional protected villages or complex terrains. This study combines deep learning and 3D modeling to assess rooftop PV potential of traditional villages in Enshi Prefecture, Hubei, China. Utilizing satellite imagery as the primary data source, we applied the U-Net model to identify usable rooftop areas. Additionally, we constructed a 3D model of the local terrain, enabling more accurate simulation of shading effects between terrains and buildings compared to existing methods. By integrating these results with solar radiation data, as well as the efficiency and performance ratios of different types of PV systems, we assessed the rooftop PV potential of four traditional villages in Enshi. The results show that the terrain in Enshi reduces the PV potential of the four traditional villages by 25.919%. By simulating the Poly-Si PV system, the study found that the annual PV potential for traditional villages in Enshi is 18,319.705322 GWh/year, with a PV efficiency of 145.021096 kWh/m2/year. This work enhances the accuracy of renewable energy development and provides a method for regional energy planning.