On the accuracy of open-source and commercial solar forecasting tools

This paper evaluates the performance of some solar forecasting tools that are available publicly over a 24-hour time horizon. These tools include one open-source, namely PVlib-python, and two commercial ones, each with different time resolutions. These tools are evaluated in terms of the forecasted direct normal irradiance (DNI) and global horizontal irradiance (GHI) which were compared to the measurements made locally, under both clear-sky, partly cloudy-sky and cloudy-sky conditions. Standardized metrics, including mean bias error, mean absolute error, mean absolute percent error, and root mean square error are used for comparison purposes. The goal is to determine how accurate these tools are for day-ahead solar forecasting, without resorting to sky imaging or ground-based irradiance monitoring.