What model do you use for power forecasting? Do you use different models for different forecast horizons e.g 1h, 1day, etc. or did you find that (recursive) multi step forecasting with the same model produces better results?
We forecast radiation just radiation data as the base for all power forecasting (conversion from solar radiation -> PV power done on the fly)
Solar radiation forecasting incorporates a few models for 1-7 days, and for now casting (0-4 hours) is based on NWP wind forecasting combined with our own cloud tracking, ML and blending with NWP models from 4-24 hours.
Some of the above might be slightly off as I’m not a meteorologist or study solar radiation modelling. Luckily Solcast founders are and we are also partnered with a project at the Australian National University working with some extremely bright people to get the science right.
If you want to get further into the details, feel free to email Nick Engerer or James Luffman (contacts on our website) about the science, or post on our community forums (forums.solcast.com.au).
