Forecasted GHI is greater than clear sky GHI

[willhobbs]

Using GFS as the forecast model, I am seeing GHI values as much as 13% higher than what ModelChain calculates when no irradiance data is provided (default ineichen clearsky model).

[cwhanse]

I’ve seen the same thing from NAM, compared to Ineichen’s clear sky model.

[wholmgren]

There are two reasons why the forecast module's irradiance data will be different from the ineichen model's:

1. The forecast module uses the Liu Jordan model to convert a cloud cover forecast to an irradiance forecast. It's going to give different results even in clear sky conditions.
2. The forecast module fails to supply the Liu Jordan model with a time-varying top-of-atmosphere irradiance. I'll fix this later today.

@cwhanse are you referring to the raw NAM data or the forecast module's processed version of it?

[cwhanse]

I’m referring to the GHI values in the NAM forecast downloaded from NOAA.

[wholmgren]

@willhobbs I wrote a new cloud cover to irradiance algorithm that's based on the ineichen function and the turbidity file. It's basically this...

solpos = location.get_solarposition(cloud_cover.index)
cs = location.get_clearsky(cloud_cover.index, model='ineichen')
# offset and cloud cover in decimal units here
ghi = (offset + (1 - offset) * (1 - cloud_cover)) * ghi_clear
dni = disc(ghi, solpos['zenith'], cloud_cover.index)['dni']
dhi = ghi - dni * np.cos(np.radians(solpos['zenith']))

See the updated documentation for more.

Lot's of room for experimentation and likely improvements. I have no idea how the accuracy of the two methods ('clearsky_scaling' and 'liujordan') compares.

I pushed those changes to both pvlib's and my own forecast branches.

[wholmgren]

Closed by #180.