Rooftop PV Potential Estimations: Automated Orthographic Satellite Image Recognition based on Publicly Available Data
This paper presents a new approach for the estimation of rooftop photovoltaic potentials. It extracts building footprints from OpenStreetMap and combines these with orthographic satellite images from Bing Maps. These images are analyzed using image recognition techniques in order to identify the ridge line as well as structures like chimneys and windows on each building’s roof. In combination with statistical assumptions about the roof’s inclination angle, the exact shape, size and orientation of the partial roof areas can be calculated for each building. For the resulting areas, an irradiance simulation is conducted and combined with a PV electricity yield model in order to calculate electricity generation profiles with a high spatial and temporal resolution. The main advantages of this approach over methods which typically employ 3D models are, that no resources are required for data acquisition and it can be applied worldwide. The method is applied to German cities and an evaluation indicates a success rate of over 70%.