Add decomposition models to allow GHI-only weather files

[williamhobbs]

Problem: SAM doesn't allow weather files that only have GHI measurements for irradiance.

Possible solution: SAM could add decomposition models to calculate DNI/DHI from GHI, allowing users use weather files that only include GHI (and temperature and wind speed).

Alternatives: Alternative options I have considered:

• It is possible for users to pre-process their weather data using pvlib or other tools to create a weather file with DHI/DNI that SAM can use. This takes additional time and requires additional software, and knowledge on how to use it. It seems to be the best current option.
• Use ground-measured data that includes DHI/DNI. Equipment to get these measurements is more expensive to purchase and maintain than GHI-only measurements. Existing publicly-available data (SURFRAD, MIDC, etc.) has limited numbers of sites.
• If the user's end goal is sub-hourly analysis, 5-min NSRDB data could be used, but there are tradeoffs: 1) it is limited to 5-min, while 1-min may be needed, 2) the relatively large area represented in each NSRDB grid cell (2 km x 2 km) could add significant smoothing of irradiance variability.

Additional context: SAM currently requires either 2 irradiance measurements (2 of: GHI, DHI, and DNI) or POA irradiance in a weather file. But some users may have access to high quality ground measurements that only include GHI*. Being able to use these ground measurements could be beneficial, particularly for things like sub-hourly analysis. Use-case examples include PV-Battery with PV smoothing dispatch and sub-hourly clipping analysis.

Regarding specific decomposition models, Erbs (used by PVsyst) and/or DIRINT (found to have lowest error of several models here: https://www.osti.gov/pages/servlets/purl/1235343) could make good options. Both are available in https://github.com/pvlib/pvlib-python (https://pvlib-python.readthedocs.io/en/stable/api.html#dni-estimation-models).

(edited after the helpful comment from @kanderso-nrel below): Since SAM already uses a version of DIRINT for decomposition in some options for POA as input, this functionality is similar to existing functionality in SAM.

* For example, from a nearby solar plant that has been operating for a year or more. These facilities typically include GHI and POA, but POA measurements could be for different tracking/orientation.

[kandersolar]

Note that SSC already has decomposition code for DISC/DIRINT in /shared/lib_irradproc.cpp, so that part might already be done

[williamhobbs]

Thanks, @kanderso-nrel. I made a small update to my original comment based on this.

[cpaulgilman]

Another request for this feature: https://sam.nrel.gov/forum/forum-general/4008