Sky imaging is used to determine local obstructions that may have an important impact on solar availability of an autonomous system powered by photovoltaic (PV) energy. However, there is currently no inexpensive, yet simple, solution that incorporates the determined horizon with weather database and system specifications for forecasting purposes. In this work, we first look at the autocalibration of cheap fisheye lens clipped onto a smartphone to extract a complex horizon from a sky image. It is then coupled with PVGIS irradiance database to generate a forecast. Finally, we contrast this result with experimental ones taken during March 2023. Overall, the shading information improves forecast precision but error still persists due to factors like weather variability and the difference between irradiance received by the solar panel's large surface and what is captured by a tiny sensor.