Evaluation of Photovoltaic Utilization Potential on Building Rooftops based on Image Recognition

In the face of escalating global energy shortages and the pressing issue of climate change, nations are increasingly prioritizing the protection of the ecological environment and the transition to sustainable renewable energy. Promoting the sustainable development of renewable resources has become a common goal. Among these resources., solar energy stands out as a pristine and sustainable option with promising prospects. Rooftop photovoltaic (PV) installations, leveraging their unrestricted geographical applicability and minimal land resource requirements, emerge as the forefront choice for harnessing solar energy. In this work, we propose an improved SegFormer model to identify different types of building roofs and assess the PV potential of these roofs, integrating solar radiation data from the targeted region. Our experiments in N anjing reveal compelling results: 412,106 building roofs are deemed suitable, covering an area of 170.6 square kilometers, with an installed capacity of 23,521.19 megawatts. The annual PV power generation capacity reaches an impressive 25,391.22 gigawatt-hours, and the annual carbon reduction potential is 14.48 million tons, indicating substantial potential for PV installations, power generation and carbon reduction. Importantly, the method proposed herein has the capacity to swiftly evaluate city-scale PV potential, offering essential data support for PV system planning within the power grid.