AbstractThis article describes the method of deriving Global Horizontal Irradiance (GHI) from combining power measurements with static meta data (tilt, orientation, brand/type) of rooftop photovoltaic (PV)‐systems. This inverse PV model implements a forward yield model that is based on a modified Orgill and Hollands decomposition model and the Perez transposition model for irradiance. The forward as well as the inverse PV model were validated with DC power measurements of four different mono‐ and polycrystalline modules combined with weather station data (2 minute resolution data over a period ranging from the 11th of June through the 24th of August 2014). The bias‐corrected forward PV model shows a best (r)RMSE of 16.0 W (15.1%) with a (r)MBE of −1.67 W (−1.57%) for one of the polycrystalline modules. The bias‐corrected inverse PV model shows a best (r)RMSE of 65.6 Wm−2 15.1% with a (r)MBE of 0.994 Wm−2 (0.229%) for one of the polycrystalline modules. Similar results were obtained for the three other modules.